© 2008 The Royal College of Psychiatrists
Area factors and suicide: 5-year follow-up of the Northern Ireland population
Epidemiology and Public Health, Centre for Clinical and Population Sciences, Queen's University, Belfast
Correspondence: Dermot O'Reilly, Epidemiology and Public Health, Centre for Clinical and Population Sciences, Mulhouse, Royal Victoria Hospital, Grosvenor Road, Belfast BT12 6BJ, UK. Email: d.oreilly{at}qub.ac.uk
None.
 TOP ABSTRACT INTRODUCTIONMethodResultsDiscussionREFERENCES |
|---|
Background
Suicide rates vary markedly between areas but it is unclear whether this is due to differences in population composition or to contextual factors operating at an area level.
Aims
To determine if area factors are independently related to suicide risk after adjustment for individual and family characteristics.
Method
A 5-year record linkage study was conducted of 1 116 748 non-institutionalised individuals aged 16–74 years, enumerated at the 2001 Northern Ireland census.
Results
The cohort experienced 566 suicides during follow-up. Suicide risks were lowest for women and for those who were married or cohabiting. Indicators of individual and household disadvantage and economic and health status at the time of the census were also strongly related to risk of suicide. The higher rates of suicide in the more deprived and socially fragmented areas disappeared after adjustment for individual and household factors. There was no significant relationship between population density and risk of suicide.
Conclusions
Differences in rates of suicide between areas are predominantly due to population characteristics rather than to area-level factors, which suggests that policies targeted at area-level factors are unlikely to significantly influence suicides rates.
|
TOPABSTRACTINTRODUCTIONMethodResultsDiscussionREFERENCES |
|---|
Research shows that suicide risk is related to both individual and household characteristics such as age, gender, marital status and socio-economic circumstances: it is less clear whether area of residence constitutes an additional, independent risk. Previous work has demonstrated higher prevalence of social problems, mental disorder and suicidal behaviour in inner cities1,2 and higher suicide rates in deprived or socially fragmented areas.3 Whitley et al also showed that suicide rates increased most in areas that recorded the largest increases in deprivation or social fragmentation over time.4 However, such studies cannot determine whether these are area effects (context) or due to the characteristics of the people living in these areas (composition). Such findings need to be confirmed using longitudinal data that include both individual and area characteristics. We report on such a study, based on a 5-year follow-up of the entire Northern Ireland population enumerated in the 2001 census.
|
TOPABSTRACTINTRODUCTIONMethodResultsDiscussionREFERENCES |
|---|
In 2006 the Northern Ireland Statistics and Research Agency (NISRA) undertook an exercise to link all those enumerated at the Northern Ireland 2001 census to deaths to residents in the subsequent 5 years. This formed a 5-year longitudinal study of the whole enumerated population, linking 94% of all deaths occurring over this period. Details of the linkage process are described elsewhere.5 These data were anonymised, held in a safe setting by NISRA and made available to the research team for this study. In keeping with established practice both definite suicides and deaths of undetermined intent were combined to define suicide (ICD–10 codes X60–X84, Y10–Y34, Y87.0),6 reducing the possible effects of misclassification. The analysis was limited to enumerated individuals aged 16–74 years at the time of the census: this allowed socio-economic factors, which in the 2001 UK census were coded only for those less than 75 years old, to be applied to the whole of the analysis population. A total of 12 055 individuals living in communal establishments (such as nursing or residential homes, hospitals and prisons) were also excluded from the analysis as they would be little influenced by area-level factors. Eight suicides (or 1.4% of all suicides occurring over the study period) were recorded for this excluded group. The resulting cohort available for analysis comprised 1 116 748 individuals.
Individual and household characteristics
All attributes of the cohort members were as described on the census
record. Marital status (categorised as married or cohabiting; never married or
single; separated or divorced; and widowed) and household size (the number of
people of any age living at the address) were included as individual-level
measures of social support. Because initial analysis showed excess risk of
suicide in single-person households but no difference in risk between
households with two or more residents, a binary household size indicator was
generated (single-person households, households with two or more people).
Socio-economic status was assessed using housing tenure (categorised as
owner-occupied, private renting or social renting); car availability (two or
more cars, one car or no car) and social class using the National Statistics
Socio-economic
Classification.7
These were then combined to derive a single measure of relative material
disadvantage comprising eight categories ranked from the least deprived group
(professional owner-occupiers with access to at least two cars), to the most
deprived (unemployed people living in socially rented accommodation with no
car access). Economic activity, known to be independently associated with
suicide, was also
included.8–11
The two census questions on self-reported health status were also included:
one asked about the presence of a limiting long-term illness and required a
`yes/no' response; the other, on general health in the year preceding the
census, offered three responses – `good', `fairly good' and `not
good'.
Area characteristics
Three indicators relating to area of residence were derived at census
super-output area level (a standard government administrative area, with
average population size 1894): material deprivation, population density and
social fragmentation. Material deprivation was defined as the proportion in
the super-output area in receipt of means-tested social security
benefits.12
Population density was measured as the census population divided by the area,
in km2; this was included as a proxy for the urban–rural
character of the area of residence. Following seminal work by
Congdon3 and Whitley
et al,4 a
measure of social fragmentation was constructed using four census variables:
the percentage of people in private rented accommodation; the percentage of
the adult population who were unmarried; the percentage of the population aged
less than 65 years who were living alone; and the percentage population
turnover in the year preceding the census. The fragmentation score was an
unweighted sum of the standardised levels of these variables, giving a mean of
0.00 (s.d.=3.27). These areas were then ranked (separately) for each set of
area characteristics and split into quintiles containing approximately equal
proportions of the population.
Statistical methods
The relationship between cohort characteristics and death due to suicide
was investigated using Cox proportional hazards modelling. Likelihood ratio
statistics were used to test for differences in hazard rates between
categories and trends across categories. Analysis was undertaken in two
stages: the first to build a model that best described the individual and
household factors associated with increased suicide risk, and the second to
determine whether or not the area characteristics contributed to suicide risk
independent of these. The Cox proportional hazard model assumes the
independence of individual responses, an assumption that might not hold if
responses within an area were correlated. Any such clustering within areas
could exaggerate the precision with which associations are estimated.
Sensitivity analyses were conducted using generalised estimating equation
models to account for any within-area
clustering.13
Generalised estimating equation logistic regression population average models
were used to make inferences concerning area
characteristics.14
Logistic regression models were used to conduct these sensitivity analyses,
both for simplicity and because the estimates from the final Cox proportional
hazard model were similar to estimates from a corresponding logistic
regression model.
|
TOPABSTRACTINTRODUCTIONMethodResultsDiscussionREFERENCES |
|---|
In the 5 years of follow-up 566 deaths were registered as either suicide or of `undetermined intent': 425 (75.1%) deaths were of men and 75.3% were of people aged less than 55 years at the time of census.
Individual and household-level indicators of suicide risk
Female suicide risk was about a third that of males (hazard ratio
(HR)=0.30, 95% CI 0.24–0.37). Table
1 shows the relationship between suicide and individual and
household factors. After adjustment for age and gender, those currently
married or cohabiting at the time of census showed the lowest risks compared
with all other marital status categories, with the excess risk associated with
the single/never married and separated/divorced categories being maintained
even after inclusion of the other demographic and socio-economic factors.
Those living alone were associated with higher suicide risk even after
adjustment for age, gender and marital status (HR=1.53, 95% CI
1.18–1.99), although further adjustment for socio-economic and general
health status weakened this association considerably (HR=1.28, 95% CI
0.98–1.67; P=0.073).
|
View this table: [in a new window] |
Table 1 Individual and household factors associated with suicide risk in people
aged 16–74 years: Cox proportional hazards analysis
|
There was a strong correlation between economic activity and deprivation, so mutual adjustment generally attenuated the effects of each. Among the economic activity categories those who were permanently ill had the highest suicide risks, with a four-fold excess after adjustment for age, gender, marital status and household size (despite constituting only 9.0% of the cohort they accounted for 25.3% of suicides). However, about half of this excess risk was explained by adjustment for disadvantage and health status. People defined as either `homemakers' (90.6% of whom were female) or `other economically inactive' had a 60% excess suicide risk when compared with those in employment. After adjustment for age, gender, marital status and household composition, unemployed people had a 68% excess suicide risk when compared with employed people (HR=1.68, 95% CI 1.20–2.35). However, further adjustment for measures of deprivation and baseline health status reduced this to a 28% excess (HR=1.28, 95% CI 0.89–1.84). There was a strong and graded relationship between individual and household deprivation and risk of suicide, although this was somewhat attenuated by the addition of health factors in the modelling. Poorer health status at the time of census, whether measured in terms of long-term illness or general health, was also strongly associated with higher suicide risk. However, when both self-reported health measures were simultaneously included in a model `general health' remained significant whereas `limiting long-term illness' became non-significant (likelihood ratio test P<0.05). For reasons of parsimony, therefore, long-term illness was not included in the final presented analysis.
Area-level indicators of suicide risk
Despite the equal distribution of population across quintiles, the more
fragmented and deprived areas contributed a disproportionate number of
suicides (Table 2). This is
reflected in the increasing gradient of HRs in the age- and gender-adjusted
models. (Note: when using the logistic regression generalised estimating
equation models that allowed adjustment for individual-level covariates and
clustering within areas, the area-level estimates of association and
corresponding significance levels were little altered from those presented in
Table 2 and have not been
shown.) After further adjustment for the individual and household factors
(marital status, household composition, socio-economic and employment status,
and baseline health status) the association between suicide risk and area
deprivation and levels of social fragmentation disappeared. Patterns
associated with population density appeared slightly more complex: although
the model adjusted for age and gender showed some indication of higher risk
for those in the least and most densely populated quintiles when compared with
the central quintiles, this disappeared in the fully adjusted model, which
showed only weak evidence of an effect associated with the least densely
populated quintile.
|
View this table: [in a new window] |
Table 2 Area measures and risk of suicide
|
|
TOPABSTRACTINTRODUCTIONMethodResultsDiscussionREFERENCES |
|---|
With a cohort of more than 1 million community residents, this is one of the largest longitudinal studies of suicide risk undertaken in the UK. It has generally confirmed the known associations between suicide and indicators of isolation and socio-economic disadvantage at both individual and household level. Married or cohabiting individuals had the lowest risk, compared with both single/never married and the separated/divorced categories. The analysis suggests that at least some of this excess is due to socio-economic disadvantage. Household composition is a factor that is not usually examined as separate from marital status, and although the increased risk associated with single-person households was non-significant in the fully adjusted model, its importance as an indication of a developing social trend suggests that it should be included in future studies. The non-significant excess risk of suicide among the unemployed requires some explanation: until fairly recently Northern Ireland consistently had the highest rates of unemployment, long-term unemployment and proportions of the working-age population defined as economically inactive of any region of the UK, and in such circumstances, where future employment chances are slim, it is thought that many of those who might otherwise register as unemployed choose to register as economically inactive owing to chronic ill health.8 This therefore may displace some of the effect of unemployment to those defined as permanently ill. Of the two self-reported morbidity measures, only general health in the previous year was significantly related to suicide risk; limiting long-term illness was not. This may be because long-term illness primarily reflects physical aspects of health status, whereas general health, being a less specific measure, may be more sensitive to psychiatric morbidity.
Area characteristics and comparison with other studies
The study showed that, after controlling for individual and household
characteristics, area of residence did not exert an independent influence on
suicide risk, suggesting that variation in suicide rates between areas is
explained by differences between the types of people living in these areas.
This can be seen as contrasting with some findings from recent ecological
studies in Great Britain and other high-income countries, and can imply that
some shift of emphasis in policy may be appropriate. For example, Congdon,
Whitley et al and Gunnell et al have shown strong
associations between rates of suicide and self-harm and levels of deprivation
or social
fragmentation,3,4,15
factors that have also been associated by Allardyce et al with
increased rates of admission for first episodes of psychosis in
Scotland.16 In the
USA, Almog et al and Curtis et al have demonstrated higher
rates of admission to psychiatric hospitals in New York from areas that were
more deprived or that had higher levels of social
fragmentation,17,18
and Hempstead showed that suicides in New Jersey tended to be higher in areas
characterised by low population density and a higher proportion of
single-person families, whereas non-fatal injuries were more closely related
to indicators of
deprivation.19
Fernquist & Cutright showed that rates of suicide in 21 high-income
countries throughout the world were strongly associated with a range of
indicators of societal
integration.20
However, ecological studies cannot determine if the variation between areas is
due to concentrations of at-risk individuals in these areas. To do this
requires studies that can examine the influence of area-level factors while
simultaneously adjusting for individual-level factors. There have been
relatively few such studies to date and none in the UK that has examined
suicide. Reijneveld & Schene in a secondary analysis of Amsterdam
residents showed that the higher prevalence of mental disorders (as assessed
by the 12-item General Health Questionnaire) in deprived areas was explained
by the higher concentration of deprived people in these
areas,21 and an
analysis of the British Household Panel Study reported only a weak association
between aggregate measures of deprivation and measures of mental health (also
assessed using the General Health Questionnaire) after adjustment for
individual
measures.22 On the
other hand, an analysis of Welsh Health Survey data by Skapinakis et
al found a significant, if small, regional effect on mental health
(measured using the mental health index of the 36-item Short Form Health
Survey), explaining less than 1% of the total
variance.24
To date there has been no other longitudinal study in the UK that has included both individual and area characteristics and that has had suicide as the end point. Hawton et al followed up patients in an ecological study relating rates of self-harm with deprivation and social fragmentation in Oxford to show that their characteristics were generally in keeping with those of the areas in which they lived,24 and Johnston et al, in a prospective analysis of self-harm in Manchester, showed that there was no association between area-level indicators of deprivation or social fragmentation and the risk of repetition of self-harm after adjustment for individual factors.25 It is difficult, however, to extrapolate from these studies, because it is recognised that the epidemiologies of self-harm and suicide are quite distinct.19 There have been two nested case–control studies based on the Danish longitudinal registries that looked at variations in suicide risk between areas while adjusting for individual factors. The first showed a higher suicide risk in urban areas among women but a lower risk for men, i.e. that the area characteristics had different effects on different subgroups of the population.26 The second study showed that the increased risk of suicide in poorer areas was greatly attenuated by controlling for differences in the people within these areas, and the study's authors concluded that the ecological associations were primarily due to the proportions of high-risk individuals living in particular areas.27 Martikainen et al, in a large study of 13 589 suicides in Finland between 1991 and 2002, found that although the area effects were greatly reduced by adjustment for individual factors, they remained significant and were more important for men than for women and for alcohol-related rather than for non-alcohol-related suicides.28
Methodological issues
This study has some methodological considerations that require comment. The
initial cohort was large enough to allow sufficient events to accrue in a
relatively short time, reducing the likelihood that the attributes of cohort
members would change over the follow-up period. Other studies have had to
aggregate up to 10 or more years of data to obtain numbers of events
sufficient to
analyse.29
Excluding people resident in communal establishments from the analysis would
also have increased the chances of demonstrating a significant area-level
effect if one truly existed. On the other hand, our study was not sufficiently
large to confidently examine variation by age and gender, and although we
formally tested and found no significant interaction between these and the
area factors, the study lacked sufficient power for us to be confident that
none truly existed.
Although the study was based on data from the 2001 census and the General Registrar's mortality records, 6.0% of all deaths could not be linked to a census record. This can arise if either the dead person was not enumerated or difficulties arise in matching death records to corresponding census forms. Although not often reported, this is a limitation on all record linkage studies, with the notable exception of those in Scandinavian countries where universal population registration systems are the norm and an almost 100% linkage is regularly achieved. It is not clear whether the non-linkage of this small proportion of deaths leads to bias (or its directionality), as separate analysis has shown the phenomenon to be more common in young adults (especially males), the unmarried and those living in the most deprived areas.5 Weich & Lewis have suggested that this might explain the generally weaker relationship between poverty and mental ill health found in longitudinal studies compared with cross-sectional studies.30 The area-level unit of analysis can be another significant source of variation and it is recognised that there can be difficulty in choosing an appropriate geographical unit to define community or neighbourhood, neither of which are likely to equate to routine administrative areas such as electoral wards.22 Additionally, in studies in which a significant area effect has been demonstrated, the effect sizes have been small,23 and may vary according to the size of the selected area unit. Kunce & Anderson, discussing the `contrariness' in the findings of ecological studies related to suicides, suggest that studies employing large aggregates of the population tend to bolster contextual influences on suicide,31 whereas data gathered on smaller aggregates yield little or weakened support for Durkheim's societal factors.32 The area-level indicators included in our study were derived from census super-output areas, smaller than either the London electoral wards used by Congdon or the 633 parliamentary constituents across Britain used by Whitley et al.3,4 This was considered appropriate, on the assumption that the smaller the population encompassed by an area the more homogeneous it is likely to be.
Finally, research related to area-level effects on health is dependent on the measures of deprivation and fragmentation. As Congdon has stated,33 deprivation measures should ideally include income and indicators of fragmentation should include measures of community ties, but in practice we are generally limited to data available from the census. It is possible that the measures of social fragmentation or deprivation, which worked so well in the ecological studies in Great Britain, are not so applicable in Northern Ireland, which contains only one city of note (Belfast) and is more rural in comparison with the rest of the UK. However, a comparison of the constituent elements of the fragmentation measure at the time of the census showed a significant difference only for population turnover (Northern Ireland 9.3%, England & Wales 12.2%),34,35 and ecological studies in Northern Ireland36 have replicated the associations between deprivation and health found in equivalent studies in the rest of the UK.37,38
Policy implications
Most ecological studies have found a strong link between area factors and
mental health and risk of suicide, and have concluded that appropriate
health-promoting and protection measures should be directed towards areas at
risk as well as towards individuals at risk. However, the conclusions from
studies such as this, which control for both individual and area-level
factors, are that most – if not all – of the variation between
areas arises because of differences between the populations within these
areas. This suggests that policies targeted at area-level factors are unlikely
to have a material impact on suicide rates.
|
TOPABSTRACTINTRODUCTIONMethodResultsDiscussionREFERENCES |
|---|
-
1
- Jarvis F, Ferrence G, Whitehead P, Johnson FG. The ecology of self-injury: a multivariate approach. Suicide Life Threat Behav 1982; 12: 90 -102.[Medline] 2
- Wing J. Epidemiologically Based Mental Health Needs Assessments: Review of Research on Psychiatric Disorders (ICD10, F2–F6). Chadwick Healey, 1982. 3
- Congdon P. Suicide and parasuicide in London: a small-area study.
Urban Stud 1996;
33: 137-58.
[Abstract/Free Full Text] 4 - Whitley E, Gunnell D, Dorling D, Davey Smith G. Ecological study of
social fragmentation, poverty, and suicide. BMJ 1999; 319: 1034
-7.
[Abstract/Free Full Text] 5 - O'Reilly D, Rosato M, Connolly S. Unlinked vital events in census based longitudinal studies can bias subsequent analysis. J Clin Epidemiol 2007; (in press). 6
- World Health Organization. International Statistical Classification of Diseases and Related Health Problems (ICD–10). WHO, 1992. 7
- Rose D, Pevalin D (eds). A Researcher's Guide to the National Statistics Socio-economic Classification. Sage, 2002 . 8
- Moser K, Fox A, Jones D. Unemployment and mortality in the OPCS Longitudinal Study. Lancet 1984; 2: 1324-9.[CrossRef][Medline] 9
- Iversen L, Andersen O, Andersen P, Christoffersen K, Keiding N.
Unemployment and mortality in Denmark, 1970–80.
BMJ 1987;
295: 879
-84.
[Abstract/Free Full Text] 10 - Martikainen P. Unemployment and mortality amongst Finnish men,
1981–5. BMJ 1990;
301: 407
-11.
[Abstract/Free Full Text] 11 - Gunnell D, Peters T, Kammerling M, Brooks J. The relation between
parasuicide, suicide, psychiatric admissions, and socioeconomic deprivation.
BMJ 1995;
311: 226
-30.
[Abstract/Free Full Text] 12 - Northern Ireland Statistics and Research Agency. Northern Ireland Multiple Deprivation Measure 2005. NISRA, 2005. 13
- Liang KY, Zeger SL. Longitudinal data analysis using generalized
linear models. Biometrika 1986;
73: 13-22.
[Abstract/Free Full Text] 14 - Hosmer DW, Lemeshow S. Applied Logistic Regression (2nd edn). Wiley, 2000. 15
- Gunnell D, Shepperd M, Evans M. Are recent changes in deliberate self-harm associated with changes in socio-economic conditions? An ecological analysis of patterns of deliberate self-harm in Bristol 1972–3 and 1995–6. Psychol Med 2000; 30: 1197 -203.[CrossRef][Medline] 16
- Allardyce J, Gilmour H, Atkinson J, Rapson T, Bishop J, McCreadie
R. Social fragmentation, deprivation and urbanicity: relation to
first-admission rates for psychoses. Br J Psychiatry 2005; 187: 401
-6.
[Abstract/Free Full Text] 17 - Almog M, Curtis S, Copeland A, Congdon P. Geographical variation in acute psychiatric admissions within New York City 1990–2000: growing inequalities in service use? Soc Sci Med 2004; 59: 361 -76.[CrossRef][Medline] 18
- Curtis S, Copeland A, Fagg J, Congdon P, Almog M, Fitzpatrick J. The ecological relationship between deprivation, social isolation and rates of hospital admission for acute psychiatric care: a comparison of London and New York City. Health Place 2006; 12: 19-37.[CrossRef][Medline] 19
- Hempstead K. The geography of self-injury: spatial patterns in attempted and completed suicide. Soc Sci Med 2006; 62: 3189 -96. 20
- Fernquist R, Cutright P. Societal integration and age-standardised suicide rates in 21 developed countries, 1955–1989. Soc Sci Res 1998; 27: 109 -27.[CrossRef][Medline] 21
- Reijneveld S, Schene A. Higher prevalence of mental disorders in socioeconomically deprived urban areas in The Netherlands: community or personal disadvantage? J Epidemiol Community Health 1998; 52: 2 -7.[Abstract] 22
- Weich S, Twigg L, Holt G, Lewis G, Jones K. Contextual risk factors
for the common mental disorders in Britain: a multilevel investigation of the
effects of place. J Epidemiol Community Health 2003; 57: 616
-21.
[Abstract/Free Full Text] 23 - Skapinakis P, Lewis G, Araya R, Jones K, Williams G. Mental health
inequalities in Wales, UK: multi-level investigation of the effect of area
deprivation. Br J Psychiatry 2005;
186: 417
-22.
[Abstract/Free Full Text] 24 - Hawton K, Harriss L, Hodder K, Simkin S, Gunnell D. The influence of the economic and social environment on deliberate self-harm and suicide: an ecological and person-based study. Psychol Med 2001; 31: 827 -36.[CrossRef][Medline] 25
- Johnston A, Cooper J, Webb R, Kapur N. Individual- and area-level
predictors of self-harm repetition. Br J Psychiatry 2006; 189: 416
-21.
[Abstract/Free Full Text] 26 - Qin P, Agerbo E, Bo Mortensen P. Suicide risk in relation to
socioeconomic, demographic, psychiatric, and familial factors: a national
register–based study of all suicides in Denmark, 1981–1997.
Am J Psychiatry 2003;
160: 765
-72.
[Abstract/Free Full Text] 27 - Agerbo E, Sterne J, Gunnell D. Combining individual and ecological data to determine compositional and contextual socio-economic risk factors for suicide. Soc Sci Med 2007; 64: 451-61.[CrossRef][Medline] 28
- Martikainen P, Mäki N, Blomgren J. The effects of area and individual social characteristics on suicide risk: a multilevel study of relative contribution and effect modification. Eur J Popul 2004; 20: 323 -50.[CrossRef] 29
- Lewis G, Sloggett A. Suicide, deprivation, and unemployment: record
linkage study. BMJ 1998;
317: 1283
-6.
[Abstract/Free Full Text] 30 - Weich S, Lewis G. Poverty, unemployment, and common mental
disorders: population based cohort study. BMJ 1998; 317: 115
-19.
[Abstract/Free Full Text] 31 - Kunce M, Anderson A. The impact of socioeconomic factors on state
suicide rates: a methodological note. Urban Stud 2002; 39: 155
-62.
[Abstract/Free Full Text] 32 - Durkheim E. Suicide: A Study in Sociology (transl. Spalding J, Simpson G). Routledge & Kegan Paul, 1952 . 33
- Congdon P. Commentary: contextual effects: index construction and
technique. Int J Epidemiol 2004;
33: 741-2.
[Free Full Text] 34 - Northern Ireland Statistics and Research Agency. Northern Ireland 2001 Output. http://www.nisra.gov.uk/census/census2001output/index.htm. 35
- Office for National Statistics. Census 2001. http://www.statistics.gov.uk/census2001/census2001.asp. 36
- Stringer P. Health Inequalities, Religious Affiliation and Urban-rural Status: A Report to the Department of Health and Social Services (NI). Northern Ireland DHSS, 1992 . 37
- Townsend P, Philimore P, Beattie A. Health and Inequalities in the North. Croom Helm, 1988 . 38
- Carstairs V, Morris R. Deprivation and Health in Scotland. Aberdeen University Press, 1991 . 39
- O'Reilly G, O'Reilly D, Rosato M, Connolly S. Urban and rural variations in morbidity and mortality in Northern Ireland. BMC Public Health 2007; 7: 123 (http://www.biomedcentral.com/1471-2458/7/123).[CrossRef][Medline]
Related articles in BJP:
- Highlights of this issue
- Kimberlie Dean
BJP 2008 192: A6.[Full Text]
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||