Neighbourhood income and mental health: A multilevel follow-up study of psychiatric hospital admissions among 4.5 million women and men
Introduction
The purpose of this multilevel follow-up study was to investigate whether there is an association between neighbourhood socioeconomic characteristics and psychiatric hospital admissions, beyond individual demographic and socioeconomic characteristics. Previous research has shown an association between individual socioeconomic characteristics and psychiatric hospital admissions, which reflects the social gradient in mental health in many industrialised countries (Keskimaki et al., 1995; Lorant et al., 2003). In addition, an ecological association between neighbourhood socioeconomic characteristics and psychiatric hospital admission rates has been found in previous research, where social deprivation was measured on the basis of small geographic areas, electoral wards, and districts (Boardman et al., 1997; Harrison et al., 1995; Koppel and McGuffin, 1999; Malmstrom et al., 1999).
The social environment shared by individuals within a neighbourhood could influence mental health beyond individual characteristics, i.e. the neighbourhood effect. However, a Dutch study of 4892 individuals, which applied multilevel models in the analyses, found that the increased prevalence of mental disorders in deprived urban areas is purely a result of a concentration of people with low socioeconomic status in such areas, i.e. the individual effect (Reijneveld and Schene, 1998). During the last decade, multilevel analyses have made it possible to separate the individual effect from the neighbourhood effect on health. Thus, individuals (level 1) nested within neighbourhoods (level 2) can be analysed with respect to the average disease risk (fixed effects) and the variance around the average disease risk (random effects) at multiple levels (Snijders and Bosker, 1999). A next step in furthering this field of research is to use multilevel models in prospective studies of large populations, after adjustment for several possible confounders at the individual level.
In this multilevel follow-up study we included individual age, gender, marital status, country of birth, income, and education of the entire Swedish population of 4.5 million women and men aged 25–64 years. Neighbourhood socioeconomic characteristics were defined based on the proportion of people with low income in the neighbourhood. In addition, small geographic units covering the whole of Sweden were used to define neighbourhood, which is an advantage especially since many other studies of the neighbourhood effect on health have used much larger geographic units. In qualitative studies, such small geographic units have been shown to be consistent with how residents themselves define their neighbourhoods (Bond Huie, 2001).
The first aim of this study was to examine whether there is an association between neighbourhood income and first psychiatric hospital admissions after adjustment for individual-level demographic and socioeconomic characteristics (fixed effects). The second aim was to examine whether there is a difference in first psychiatric hospital admissions between neighbourhoods, measured as the between-neighbourhood variance, and how much of the total variance is at the neighbourhood level (random effects).
Section snippets
Materials and methods
This follow-up study included the entire Swedish population aged 25–64 years on 31 December 1997, a total of 2 283 157 women and 2 354 837 men. The individuals were followed from 1 January 1998 until first hospital admission for mental disorders, death from all causes, emigration or censoring on 31 December 1999. We defined first hospital admissions as first admissions during the study period. Data were obtained from a database at the Centre of Family Medicine, Karolinska Institute, containing
Results
Table 1 shows the distribution of the population by neighbourhood-level and individual-level characteristics. The population is divided by neighbourhood income into quintiles according to the proportions of individuals with low income.
Table 2 shows the age-standardised hospital admission rates for first psychiatric hospital admissions per 100 000 person-years, stratified by neighbourhood income in quintiles. There were in total 10 795 events of first psychiatric hospital admissions for women and
Discussion
The main findings of this follow-up study of 4.5 million women and men were that low neighbourhood income is associated with an increased risk of being hospitalised for mental disorder, after adjustment for individual age, marital status, country of birth, income and education. The between-neighbourhood variance indicated statistically significant differences in psychiatric hospital admissions between neighbourhoods after adjustment for the individual and neighbourhood-level variables.
Several
Conclusions
Neighbourhood socioeconomic characteristics, such as the proportions of people with low income in the neighbourhood, should be taken into account in the distribution of psychiatric health care resources. Municipally driven improvements of the social and physical environment should especially focus on low-income neighbourhoods in order to enhance people's well-being and mental health in such neighbourhoods.
Acknowledgements
This work was supported by grants from the National Institutes of Health (1R01 HL71084-01), the Swedish Council for Working Life and Social Research (2001-2373), and the Swedish Research Council to Dr Kristina Sundquist (K2005-27X-15428-01A). The authors wish to thank Sanna Sundquist, student at University of California, San Diego, for technical assistance.
References (34)
- et al.
Socioeconomic equity in Finnish hospital care in relation to need
Social Science & Medicine
(1995) Social network and social support deficits in depressed patients: a result of distorted perception?
European Archives of Psychiatry and Clinical Neuroscience
(1991)- et al.
Social indicators and the prediction of psychiatric admission in different diagnostic groups
British Journal of Psychiatry
(1997) The concept of neighborhood in health and mortality research
Sociological Spectrum
(2001)- et al.(1980)
- et al.
Evidence-based policies and practice
Environmental Planning
(2001) - et al.
Empirical comparisons of proportional hazards, poisson, and logistic regression modeling of occupational cohort data
American Journal of Industrial Medicine
(1998) - et al.
Psychosocial and psychiatric risk factors for suicide. Case-control psychological autopsy study
British Journal of Psychiatry
(2000) - Dalgard, O., 1980. Bomiljø og psykisk helse (Environment and mental health). Dissertation in Norweigan with English...
- Goldstein, H., Browne, W., Rasbash, J., 2002. Partitioning variation in multilevel models,...