OUP user menu

Collective resources or local social inequalities? Examining the social determinants of mental health in rural areas

Mylène Riva, Clare Bambra, Sarah Curtis, Lise Gauvin
DOI: http://dx.doi.org/10.1093/eurpub/ckq064 197-203 First published online: 18 May 2010

Abstract

Background: In England, although some studies report better health among rural populations, few have examined social inequalities in health within rural areas and how they compare to inequalities observed in urban settings. The objectives of this study are to examine (i) whether living in rural, in more affluent and in more socially cohesive areas is associated with better mental health; and (ii) whether being in employment is more protective for mental health in rural than in urban areas. Methods: Data on common mental disorders (CMD) and socio-demographic characteristics of 12 962 adults are from the Health Survey for England. Individuals resided in 892 areas categorized as urban or rural. Area deprivation is measured using the employment deprivation domain from the 2004 Index of Multiple Deprivation. Area social cohesion is derived from individuals’ perceptions using ecometric procedures. Data are analysed using multilevel logistic models. Results: Living in rural areas is significantly associated with lower risk of reporting CMD [odds ratio (OR): 0.81; 95% confidence interval (CI): 0.71–0.92], independently of individuals’ characteristics, and of area deprivation and social cohesion. The mental health advantage of being in employment is more important in rural areas (OR: 0.74; 95% CI: 0.58–0.95) than in urban settings. Conclusion: Living in rural areas is associated with better overall mental health. Yet inequalities in mental health between people in the workforce and those who are not are more important in rural settings. More studies are needed to understand the patterning of social inequalities in health in rural communities.

  • employment
  • mental health
  • rural health
  • small-area analysis

Introduction

In the UK, social inequalities in health in rural communities have received less attention than urban health inequalities. This can be partly explained by the considerable inequalities in health often observed between more deprived and affluent areas within the same city. Additionally, average health in the countryside appears to be better: premature mortality and illnesses are less prevalent,1,2 rural dwellers are more likely to enjoy better mental health on some measures3 and less (or equally) likely to rate their health as fair or poor.4,5 Despite this ‘rural health advantage’, widening health inequalities are observed in rural areas of England6 and of Scotland,7 and some studies report significant social inequalities in health between individuals living in rural areas.5 For mortality from suicide8,9 and from unintentional injury,10 the prevalence is relatively high in rural areas. These studies are indicative of health differences between rural and urban areas in the UK, but do not inform if there are social inequalities in health ‘within’ rural areas and whether they are of similar magnitude to those observed in urban areas. Contributing to this evidence base has the potential to influence social policy for targeting actions towards sub-groups of the population and areas which may benefit more from health service and labour market interventions.

Stafford and Marmot have proposed two frameworks to understand how area-level conditions may interact with individuals’ socioeconomic circumstances to influence health: the ‘collective resources’ and the ‘local social inequality’ models.11 Although the authors focused on socioeconomic conditions at the area and individual levels, it is possible to adapt these frameworks to examine how living in rural environments in England may differently affect the health of people with differing socioeconomic status.

The collective resources model suggests that people living in areas characterized by more and better social and material collective resources enjoy better health than other groups living elsewhere.11 These collective resources, e.g. higher quality services and amenities, wealth, employment opportunities and social support, provide opportunities for people to live healthy lives and may be especially important for poorer people as they may be more reliant on local resources and services. Over and above individuals’ characteristics, area-level deprivation and social cohesion have been reported as important determinants of various health outcomes,12 including mental health.13,14 In England, rural areas are, on average, less deprived and may be more socially cohesive than urban areas.15 These ‘collective resources’ might explain the health advantage observed in rural areas.

Disparities between individuals’ socioeconomic position and the socioeconomic conditions of their local area may interact to influence health, as suggested by the local social inequality model.11 Poorer individuals living in wealthy areas with high-quality collective resources may have worse health than poorer individuals living in more deprived communities because, for example, of limitations in purchasing goods, participating in community life, and ‘unhealthy’ effects of comparing oneself to neighbours.11 In contrast, more well-off individuals living in deprived areas may enjoy better health then their counterparts living in more affluent areas. In rural England, the socioeconomic profile of rural populations is heterogeneous, often characterized by wealthier individuals living in close proximity to people struggling to make ends meet.16 The overall rural health advantage may mask more local social inequalities in health.

Testing these frameworks empirically, Stafford and Marmot observed that individual employment grade and neighbourhood deprivation were directly associated with ill-health, but the interaction between individual poverty/wealth and area deprivation/affluence was not significant. The authors11 call for “larger studies to investigate whether the health of those in higher socioeconomic positions is ‘protected’ from the health-damaging effect of living in more deprived areas” (p.363). To our knowledge, few studies have addressed this question in relation to mental health (although some studies report interaction between area deprivation and economic inactivity at the individual level in influencing mental health outcomes),17,18 or have examined whether higher individuals’ socioeconomic position is more protective for health in rural vs. urban areas.

The aim of the study is to test the collective resources and the local social inequality models for their relevance to understand urban/rural inequalities in mental health England. Mental health problems are an important public health burden associated with reduced physical and social functioning, social exclusion and loss of productivity.19 We first examine whether living in rural, in more affluent and in more socially cohesive areas is associated with better mental health outcomes. Second, we examine whether being in employment (an important social determinant of health, including mental health outcomes)20 is more protective for mental health in rural areas, where access to better quality jobs is more limited than in urban areas. The influence of rurality on mental health is hypothesized to be explained in terms of the interplay between both models, independently of individual’s socio-demographic characteristics.

Methods

Individual sample and data

Individual data are from the Health Survey for England (HSE), an annual representative cross-sectional survey of the English population.21 Our use of the HSE data was reviewed and approved by the UK National Centre for Social Research, which linked individual data for the years 2001–03 to categorical data for small areas (these years correspond most closely to the years for which area data on deprivation were collated). The 2001–03 pooled individual sample was limited to working age adults (18–64 years of age), producing an overall study sample of 22 850 respondents.

In the HSE, mental health is assessed using the well-validated 12-item General Health Questionnaire (GHQ-12). GHQ-12 scores were dichotomized by contrasting respondents having a score of 3 or more, i.e. reporting signs of common mental disorders (CMD)—anxiety and/or depression,3,22 vs. those who did not.

Because of greater population heterogeneity characterizing rural communities,16 adjusting for the composition of areas, i.e. individuals’ characteristics, is important. Individual socio-demographic characteristics considered were age (in 10-year age groups), sex, household income (adjusted for the number of persons in the household), work status, i.e. being in employment vs. unemployed, retired, or other economically inactive, having access to a car, self-reported ethnic group (being British, Scottish, Irish, Welsh vs. other ethnic group). Models were also adjusted for social support from family and friends, as this has been identified as an important determinant of mental health.23

Area sample and data

Areas were defined using Middle Layer Super Output Areas (MSOAs) which are aggregates of Super Output Areas.24 Compared to electoral wards, MSOAs are more consistent in geographical extent and population (e.g. MSOAs contain between 5000 and 7200 people, whereas the population of electoral wards vary from fewer than 100 to more than 30 000 residents).24 Individuals from our sample resided in 3131 MSOAs.

Rurality was defined using the Department for Environment, Food, and Rural Affairs 2001 classification.25 Urban areas were classified as those encompassing settlements ≥10 000 residents and rural areas are those comprising <10 000 residents and include small towns and fringe, villages, hamlets and isolated dwellings. Because of small rural sample size, it was impossible to distinguish between different types of rural areas.

Deprivation was measured using the ‘employment deprivation domain’ of the 2004 Index of Multiple Deprivation,26 which indicates the level of economic inactivity at the area-level; it is conceptualised as “the involuntary exclusion of the working age population from the world of work” (p. 21). Six indicators compose this domain, combining information on unemployment and claimants of different incapacity benefit and welfare programs. This domain was selected because interpreting the influence on health of a single dimension of deprivation is more meaningful; it also relates more directly to the individual-level variable of work status on which this study focuses. To protect confidentiality of the HSE respondents to whom the data would be linked, MSOA data on deprivation (as released to us by the UK National Centre for Social Research) were classified into deciles. In view of small numbers of MSOAs in some deciles, MSOAs were categorized as being in the lower (1–3), middle (4–7) and higher (8–10) deciles of employment deprivation.

In the HSE, respondents report on four items relating to the social context of their local area: neighbours looking after each other, perceived trust, helpfulness and fairness of neighbours. Data on perceptions, adjusted for selected individuals characteristics, were aggregated at the MSOA level by means of ecometrics, which is a specific type of multilevel analysis.27–29 (see Appendix 1 for detailed information on the ecometric procedure). The third-level Bayesian estimates from this model correspond to the unbiased estimation of social cohesion at the area level, from which inter-observers (level 2) and inter-items (level 1) variances are removed. Social cohesion was categorized in tertiles.

In comparison to urban areas, rural communities are significantly (P < 0.001) less likely to be characterized by employment deprivation (β = –0.27; Se = 0.034) and more likely to be characterized by higher levels of social cohesion (β = 0.28; Se = 0.033).

Statistical analysis

Data were analysed using logistic multilevel models. The collective resources model was first tested by examining the direct association between living in rural (vs. urban) areas and CMD. Then, association between area deprivation and social cohesion were modelled to examine whether more favourable material and social conditions of local areas were associated with mental health. To test the local social inequality model, the second multilevel model built on the previous one to examine whether being in work is more protective for mental health in rural communities, and also in more deprived areas; these associations were modelled as cross-level interactions. Both models were adjusted for individuals’ characteristics.

Results

The HSE sample design resulted in a sample of individuals which was not stratified within MSOAs; the numbers of individuals within each MSOA ranged from 1 to 51 (mean = 8; median = 7). Because some estimates might have been less reliable due to small individual sample size in some MSOAs, analyses were conducted on a sub-sample for which there was a minimum of 10 respondents per MSOA. This considerably reduced the sample size but provided a more robust sample of 12 962 individuals living in one of 892 MSOAs (analyses on the full sample were also conducted; strength and significance of associations were similar). Description of the full and reduced individual and area samples is presented in table 1. In the reduced sample, 22.3% people lived in rural areas (77.7% in urban areas). The proportion of people in employment is 73.9% in rural and 70.5% in urban areas. Of people in employment, 33.6%, 42.4% and 24.1% lived in respectively lower, middle and higher deciles of area-level employment deprivation.

View this table:
Table 1

Characteristics (reduced and full samples) of individuals participating in the Health Survey for England between 2001 and 2003 and of the Middle Layer Super Output Areas (MSOAs) within which they were living

Reduced sample (n = 12 962)Full sample (n = 22 850)
N%N%
Individual-level characteristics
Presence of common mental disorders
    No10 43580.518 25779.9
    Yes237518.3429718.8
    Missing1521.22961.3
Sex
    Men589645.510 26944.9
    Women706654.512 58155.1
Age (years)
    18–24225617.4403617.7
    25–34242418.7431218.9
    35–44311124.0551224.1
    45–54271220.9463420.3
    55–64245919.0435619.1
Self-reported ethnic group
    British, Scottish, Irish, Welsh11 73490.520 29888.8
    Other12159.425311.1
    Missing130.1210.1
Having access to a car
    Yes11 21686.519 44685.1
    No174213.4339614.9
    Missing40.0380.04
Social support from family and frienda
    Good801861.914 00661.3
    Medium325025.1573925.1
    Low158212.2289512.7
    Missing1120.92100.9
Work status
    In employment922571.216 13970.7
    Unemployed, retired, other economically inactiveb372328.7668229.3
    Missing140.1290.1
McClemens household score for equivalized incomeMeanSDMeanSD
1.270.421.250.41
Area characteristicsReduced sample (n = 862)Full sample (n = 3131)
Rurality
    Urban70879.4259282.8
    Rural18420.653917.2
Employment deprivation
    Lower deciles26930.288528.3
    Middle deciles37441.9126840.5
    Higher deciles24927.997831.2
  • a: Social support was assessed using the following questions: “There are people I know amongst my family or friends who: do things to make me happy; make me feel loved; can be relied on no matter what; would see that I am taken care of if I needed to be; who accept me just as I am; make me feel an important part of their lives, give me support and encouragement”.

  • b: In the sample, 5.1% were unemployed, 5.5% were retired and 18.1% were categorized as ‘other economically inactive’; this comprised people going to school full time, waiting to take up paid work already obtained, looking for paid work or a Government training scheme, intending to look for work but prevented by temporary sickness or injury, permanently unable to work because of long-term sickness or disability, looking after the home or family.21

The unconditional model revealed statistically significant between-area variation in the likelihood of reporting CMD. Across all areas, the average probability of people reporting CMD was 18.5%. Yet the plausible value range30 indicates that in some areas only 11.3% of people reported CMD, whereas in others this proportion reached 28.7% [χ2(df:891) = 1047; P < 0.001].

There were significant socio-demographic inequalities in CMD between respondents (table 2). Women were more likely than men to report CMD, whereas adults aged between 55 and 64 were less likely to report CMD in comparison to younger adults. Having better social support, being in employment and higher household income were significantly associated with lower odds of reporting CMD. The mental health advantage associated with being in employment was significantly more important than higher household income (P < 0.001). Controlling for individual characteristics of respondents did not explain all of the variation in CMD between areas which, although attenuated, remained significant.

View this table:
Table 2

Results of multilevel models testing the collective resources model to explain variation in common mental disorders among 12 962 working age adults residing in 892 local areas

OR95% CI
Intercept0.880.69–1.12
Individual characteristics
Sex
    Men1.00
    Women1.411.28–1.56
Age
    18–241.00
    25–340.860.73 –1.00
    35–440.970.84–1.12
    45–541.000.86–1.16
    55–640.620.52–0.74
Self-reported ethnic group
    Other1.00
    British, Scottish, Irish, Welsh1.080.92–1.26
Having access to a car
    No1.00
    Yes0.880.77–1.01
In employment
    No1.00
    Yes0.530.48–0.58
Household income
    Lower tertile1.00
    Middle tertile0.870.77–0.98
    Higher tertile0.860.76–0.98
Social support from family/friends
    Low1.00
    Medium0.610.53–0.70
    Good0.350.30–0.39
Area characteristics
Rurality
    Urban1.00
    Rural0.840.74–0.96
Employment deprivation (deciles)
    High deprivation1.00
    Middle deprivation0.980.87–1.10
    Low deprivation0.940.82–1.08
Social cohesion (tertiles)
    Low1.00
    Middle0.880.78–0.99
    High0.930.82–1.06
Between-area variation
    Variance component (P-value)0.05 (0.038)

Testing the collective resources model showed that living in rural areas was significantly associated with lower likelihood of reporting mental health problems (OR: 0.81; 95% CI: 0.72–0.92; data not shown), independently of individuals’ characteristics. This association remained significant when the model was adjusted for area-level deprivation and social cohesion; the fully adjusted collective resources model is presented in table 2. Among this sample, living in rural areas was significantly associated with a lower likelihood of reporting CMD. The association of CMD with employment deprivation in the local area was not statistically significant but living in areas characterized by average, but not high, social cohesion was associated with lower odds of mental health problems.

Results for the local social inequality model are shown in figure 1. Being in employment was significantly more protective for CMD in rural areas than in urban settings (OR: 0.74: 95% CI: 0.58–0.95), and in areas characterized by middle (OR: 0.74; 95%CI: 0.56–0.91) and higher levels (OR: 0.67; 95%CI: 0.51–0.87) of employment deprivation in comparison to less deprived areas. Inequalities in CMD between those in employment and those unemployed and economically inactive appeared more marked in rural and in more deprived areas, as indicated by a steeper ‘gradient’ in these areas between individuals of contrasting economic status. The mental health advantage associated with being in employment was similar for rural dwellers as for people in the most deprived areas compared with other parts of the country as a whole. Adjusting the model for these cross-level interactions did not however explain all of the variation in mental health between areas.

Figure 1

Results of multilevel models testing the local inequality model to explain in variation in common mental disorders among 12 962 working age adults residing in 892 local areas. The model is adjusted for age, sex, household income, having access to a car, self-reported ethnic group and informal social support from family and friends

Discussion

The aim of the study was to test the collective resources and the local social inequality models for their relevance to understand urban/rural inequalities in mental health in England and to examine the mental health advantage associated with being in employment in rural areas. Results show that, although rural context has a direct and positive influence on mental health, inequalities in mental health associated with being in the workforce are more important in rural communities than in urban settings.

As observed elsewhere, in comparison to urban settings living in rural areas is generally associated with better mental health, independently of individuals’ socioeconomic conditions.3 Health inequalities between rural and urban areas may reflect differences in social and material conditions of local areas, an argument consistent with the collective resources model. However, the relationship of deprivation to CMD was non-significant, while living in more socially cohesive areas was associated with lower likelihood of reporting mental health problems. Our findings extend other research conducted in the UK reporting worse mental health outcomes in deprived localities, but only for economically inactive individuals,17,18 and with better mental health outcomes in neighbourhoods with higher levels of social capital.31 Yet area deprivation and social cohesion did not ‘explain away’ the effect of rurality on CMD which remained significant, suggesting an independent influence of rural environments on population mental health.

The significant influence on mental health of the interaction between rural context and work status at the individual level supports the local social inequality model. The lower risk of CMD associated with being in employment is more important for rural than urban dwellers, suggesting that social inequalities in mental health associated with work status are more pronounced in rural communities. In addition, being employed is as protective for mental health in rural areas as it is in the most deprived areas of the country. Being in employment may provide more material resources to individuals, e.g. disposable income, and sense of purpose and meaning associated with work and protection from stressors in the local environment. This may be especially important in rural areas where employment opportunities and better quality jobs are few. A psychosocial explanation would suggest that economically active individuals living in communities with struggling labour markets compare themselves positively to their neighbours excluded from the labour force, and this has a beneficial influence on their mental health. The cross-sectional nature of the data indicates that another possible interpretation is that in rural areas (as well as the most deprived areas nationally) people with mental health problems are especially likely to become unemployed, which would be consistent with greater precarity of employment in economically marginal areas.

Considering interactions between socioeconomic characteristics at the individual level and the rural context may account for the relative wealth and health advantage of rural populations. Longitudinal and quasi-experimental studies (e.g. evaluation of naturally occurring experiments such as rural development schemes) are needed to understand and ascribe causation about how rurality and deprivation at the individual and area levels interact to influence health and health inequalities in the English countryside. In conjunction, efforts are needed to improve methodologies and validity of measurements of social and economic conditions in rural settings. More in-depth studies should be conducted to understand the pathways through which collective resources and local social inequalities influence mental health.

Results reported should be interpreted in light of some limitations. As noted above, the cross-sectional design of the study prevents analysing processes of causation linking area contexts to health, and health selection into un/employment cannot be ruled out. In rural areas, there is a problem of hidden unemployment as rural people may not identify themselves as unemployed and there are marked seasonal variations in employment.32 Numbers of people in employment may have been overestimated in the HSE sample. A dichotomous definition of rurality, contrasting urban and rural settings may mask significant variation in health associated with specific rural contexts. Larger studies investigating the complex process linking place to health in rural settings are needed.

Implications for public health research and policy

There is a strong theoretical basis for expecting area regeneration programmes which increase economic activity to be beneficial for health.33 Yet, there is little evidence to suggest that in practice, regeneration and job creation improve physical or mental health outcomes.34,35 Although tackling unemployment and economic inactivity in urban deprived areas has been the focus of regeneration initiatives (as reviewed in reference 36), it has received less attention in rural areas. Similarly, whilst there is a large body of research which suggests that work is generally good for health,37 the type and quality of work is an important mediator of this relationship.38 Low skill, low pay, high stress, insecure jobs may actually be worse for health than economic inactivity.39 These types of jobs may be more characteristic of rural labour markets. Future studies are needed to understand the patterning of employment and health not only in rural areas, but also in contrasting rural labour markets. Evaluation of interventions to tackle unemployment in rural settings and their impact on population health are wanting.

Acknowledgements

The authors would like to thank people at the National Centre for Social Research for their help in building the database and one anonymous referee for the helpful comments. An earlier version of this paper was presented at the 12th International Congress of the International Federation of Psychiatric Epidemiology, Vienna, Austria 16th-19th April 2009. LG holds a Canadian Institutes of Health Research/Centre de recherche en prévention de l′obésité, Applied Public Health Chair on Neighborhoods, Lifestyle, and Healthy Body Weight.

Conflicts of interest: None declared.

Key points

  • Living in rural areas is associated with overall better mental health, independently of individuals’ characteristics, and deprivation and social cohesion at the area level.

  • Inequalities in mental health between people in the workforce and those who are not are more important in rural settings. Being in employment is as ‘protective’ for mental health in rural areas as it is in deprived areas across England.

  • Longitudinal and quasi-experimental studies are needed to understand and explain about how rurality and deprivation at the individual and area levels interact to influence health and health inequalities in the English countryside.

  • Further studies of the health impacts of interventions to address unemployment and economic inactivity in rural areas seem to be justified by our results.

Appendix 1: ecometrics procedures

In the Health Survey for England, the social context of local areas is explored by asking respondents to report on the following four items (possible answers in bracket): (i) this area is a place where neighbours look after each other (strongly agree, agree, disagree, strongly disagree); (ii) generally speaking, would you say that most people can be trusted or you can’t be too careful in dealing with people (can be trusted, can’t be too careful, don’t know); (iii) would you say that most of the time people try to be helpful or just look out for themselves (try to be helpful, look out for themselves, don’t know); and (iv) do you think most people would take advantage of you if they got the chance or would they try to be fair (take advantage, try to be fair, don’t know). It was hypothesized that the four items are indicators of a latent construct of social cohesion in the local areas.

In a three-level multilevel model, items (level 1; n = 104 233) were conceptualized as being nested within respondents (level 2; n = 12 920), who in turn were nested within MSOAs (level 3; n = 892). The latent construct of social cohesion and observations on the four items were dichotomized (and reverse-coded where appropriate) to capture their presence/absence in the latent construct (people answering ‘don’t know’ were excluded), and were modelled simultaneously as level-1 predictors. The item ‘most people try to be helpful’ was selected as the reference category because it corresponds best to the latent construct of social cohesion.

The model was then adjusted for the following characteristics of individuals (level-2 predictors) which may influence the perceptions one holds about the context of their local area (potential reporting bias): age, sex, household income, employment status, self-reported ethnic group, self-rated health and years lived in current local area. No variables were specified at the MSOA level.

The modelling equation is as follows:

Level-1 model:Prob(Y = 1| β) = P
log[P/(1 – P)] = P0 + P1*(neighbours look after each other) + P2*(most people can be trusted) + P3*(most people would take advantage)
Level-2 model:P0 = β00 + β01*(sex) + β02*(age) + β03*(household income) + β04*(employed) + β05*(self-reported ethnic group) + β06*(more than 5 years of residence in local area) + β07*(good/very good self-rated health) + r0
P1 = β10; P2 = β20; P3 = β30
Level-3 model:β00 = γ000 + μ00
β01 = γ010 […] β07 = γ070
β10 = γ100 […] β30 = γ300

The third-level Bayesian estimates from these models correspond to the unbiased estimation of social cohesion at the area level, from which inter-observers and inter-item variances are removed. Bayesian estimates were categorized in tertiles of social cohesion.

The ecometric approach also allows determining the reliability of the area-level measure, which is calculated as the ratio of the ‘true’ score variance to the observed score variance in the sample area mean, with values ranging from 0 to 1.28 The reliability score is indicative of the average degree of ‘consensus’ on perceptions among respondents living within the same MSOA. It will be high (close to 1) when: (i) the area means vary considerably across areas or (ii) the sample size per area is large. The reliability of our measure of social cohesion at the MSOA-level was 0.451. This modest reliability could be explain by small sample size in some MSOAs (range = 9–46; mean = 14; median = 13) (e.g. Raudenbush and Sampson have established that a sample size of 25–30 will maximize area reliability).27 Another reason could be that the items used to measure the social cohesion of local area in the HSE, and the limited responses possible to determine these items (on three out of the four items, respondents had the choice between only two possible answers), might have influenced the reliability score.

References

View Abstract