6 Society, place and space Jane Boydell and Kwame McKenzie Introduction This chapter will discuss the impact of society, place and space on the incidence of psychosis. It will briefly introduce the history of social causation theory before using the well established effect of urban residence on the incidence of psychosis in general, and schizophrenia in particular, as a lens through which to consider the possibilities for, and problems with, this research field. The reasons why the socioenvironmental context seems likely to be important will be described. Recent attention has focused on the idea that neighbourhood factors might exert an effect beyond their individual equivalents. For example, the social cohesion of a neigh- bourhood might have an effect on rates of psychosis above and beyond that of individual social networks. These possibilities, and the challenges associated with them, will be discussed. History Prior to the rise of modern medicine, the cause of disease was attributed to a variety of spiritual or mechanical factors, such as the elements, humours or miasma – bad air arising out of dirt and decaying organic matter. Early public health research, built on these theories, took the environment, in particular poor areas, to be aetiologically relevant. Risk was related to place; populations, rather than individuals, were considered more vulnerable because of where they lived rather than because of their own behaviour. Pioneers of public health in the mid-nineteenth century targeted sanitation of the slums, not education on personal hygiene, considering this to be the most important way of improving health (Porter, 1997, p. 411). This ecological approach to health, however, was soon undermined by the development of new concepts; of particular importance was the discovery of disease-causing micro-organisms, and the rise of germ theory (Porter, 1997, p. 415). The importance of these developments was twofold: first, they changed Society and Psychosis, ed. Craig Morgan, Kwame McKenzie and Paul Fearon. Published by Cambridge University Press. # Cambridge University Press 2008. the target of investigation from the community as a whole to pathological agents in the community; and second, they moved the place for investigation of aetiology away from the community and into the laboratory (and the individual). Germ theory was useful, but had its problems. For instance, not everybody exposed to infection contracts a disease. Exposure, though necessary, is not always sufficient to produce illness. Because of this and other deficiencies, the so-called ‘epidemiological triangle’ approach was developed. This renewed interest in the environment but still did not bring it back centrally into the causal pathway. It posited that disease is the product of an interaction between an agent, the host and the environment. The host and environment determine exposure and susceptibility. An example would be methicillin-resistant Staphylococcus aureas (MRSA) infection, which is a current scourge of UK hospitals. The environment has a part to play, in that poor hospital hygiene increases the chance of infection being spread. However, the weakest people in a hospital, whose immune systems are not functioning fully (e.g., post-operative patients, the old and infirm), are most likely to develop an infection. The risk of MRSA infection, therefore, depends on the presence of the bug in the hospital, the hospital environment and individual vulnerability. Although the epidemiological triangle works well for infectious disease, it is more difficult to apply to chronic illnesses, such as psychotic disorders. The problem is that there is often no specific agent or exposure. There is a web of causation. Disorders develop through the complex interactions of many factors over time, which form interlocking chains of events (Krieger, 1994). Here, the environment is important, but only as a site for risk factors or behaviours that increase risk. The general move has been towards the individualisation of risk and away from the ecological basis of risk that was the genesis of public health. However, some theories have continued to emphasise the importance of society and community. They have thrived because there are significant differences in rates of illness between groups, which are not explained by known risk factors (e.g., residents in urban and rural areas), and because some social groups seem to be at increased risk for a number of disparate disorders (e.g., migrants). There are some general theories of susceptibility that do not identify single or even multiple risk factors associated with specific disorders, but seek to understand why some social groups are generally more at risk than others for a range of illnesses. One incarnation is the fundamental social cause hypothesis (Link and Phelan, 1996). This attempts to explain why disparities in illness remain between socioeconomic groups despite adequate public health interventions. In the words of Link and Phelan (1995, p. 80), ‘ .social factors such as socioeconomic status and social support are likely ‘fundamental causes’ of disease .because they embody access to important resources, affect multiple disease outcomes through multiple mechanisms, and consequently maintain an association with disease even when intervening mechanisms change.’ 78 J. Boydell and K. McKenzie In the nineteenth century, the higher mortality and morbidity rates in the lower social classes were primarily due to infectious disease. Public health interventions, such as sanitation and vaccination, had a significant impact on disparities resulting from infections, but the differences in life expectancy between the social classes remained. Over time, the reasons for these disparities were no longer infectious disease, but became chronic diseases, such as cardiac problems (Link and Phelan, 1996). The theory of fundamental social causes aims to identify factors, such as education, access to prevention messages and access to healthcare, that work together to produce health disparities. Such fundamental causes are lodged in the fabric of society rather than in the individual. This theory has heralded a recent return to ecological considerations of aetiology (Diez-Roux, 1998). A further move in this direction has been the social capital literature (McKenzie and Harpham, 2006). Social capital is a way of understanding communities and how their structure and function affect rates of mental illness. It is a group of concepts that includes individual and ecological factors. These may prove to be powerful predic- tors of rates of mental illness and outcome, but the concepts and methodologies are still under revision and there is significant work needed before any conclusions can be drawn (McKenzie and Harpham, 2006, pp. 151–7). What is exciting is that the use of multilevel modelling techniques has allowed social capital to be measured at ecological and individual levels in the same data sets. This has made it possible to disentangle the impact of the social environment on the individual from the impact on the group (McKenzie and Harpham, 2006, pp. 86–109). Conceptualising society, place and space Over time, society, place and space have been conceptualised in a number of different ways (see Chapter 4). The environment has been considered as a cause of illness at a population level and at an individual level, and as a vector of risk factors at an individual level. Psychiatrists and social scientists have a long history of interest in the impact of society, place and space on the aetiology of mental illness. For instance, Durkheim (1951) developed theories that linked social structure and suicide (see Chapter 3); Faris and Dunham (1939) investigated the social organisation of cities and mental health (see Chapter 4); and Leighton (1982), Freeman (1994) and others (e.g., McKenzie et al., 2002) have been interested in the impact of societal and socio- cultural change on mental health. There have been a number of attempts to describe society and social structure. Economic variables have been considered significant. They have included concepts of wealth, income inequality and poverty; both absolute poverty (money needed to sustain life) and relative poverty (the money needed to be able to have a normal life 79 Society, place and space within a society). Many types of inequality cluster together – the financially poor are often disadvantaged in accessing high quality education, decent housing, and so on. These socially structured inequalities have been studied through concepts such as social class, poverty and social exclusion. Socioenvironmental stressors, such as crowding, crime and fear, have been less well conceptualised and studied, at least in relation to mental illness. Social context is considered so important in the understanding of symptoms and illness that it has been written into psychopathological definitions and diag- noses. But it is often unclear what that context is; whether it should be considered as acting on individuals or whether it should be considered to be truly ecological (i.e., acting at a group level). Because of this, it is difficult to know how far this understanding helps us in developing healthy communities. A classic example of this is urbanisation. On a population level, the most important risk factor for psychosis is being born and brought up in a city. We will use this as a model for the investigation of how society, place and space may affect aetiology. Methodological issues To understand theories of how urbanisation may cause mental illness, one has to accept some simple premises. The first is that psychosis often develops over a life course, with a balance of propsychotic (i.e., risk-increasing) and antipsychotic (i.e., risk-reducing) factors moving individuals up or down a scale of risk (Figure 6.1). Propsychotic and antipsychotic risk factors are complex and exist at a number of discrete levels. For instance, a non-exhaustive list of levels would be molecular, genetic, individual, interpersonal and environmental (Figure 6.2). Investigation of the cause of psychosis is complicated by the fact that each of these levels is No psychotic symptoms Prepsychosis Psychotic symptoms Psychotic symptoms + impairment Psychosis diagnosis Propsychotic factors Antipsychotic factors Figure 6.1 Life course journey to a diagnosis of psychosis 80 J. Boydell and K. McKenzie governed by different scientific rules and so different tools are needed for assess- ment (Susser and Susser, 1996). It is perhaps easier to understand this by considering the case of smoking and premature death. Those who smoke have an increased risk of developing a variety of cancers and cardiovascular disease. The rates of illness in groups who smoke are increased. To build an evidence-based strategy to decrease the impacts of smoking on health, we may want initially to investigate the mechanisms through which smoking has its effects. The mechanisms can be interrogated in a number of ways and at a number of levels, as follows: * Molecular level How do nicotine and tar affect the contents of a cell? * Metabolic level How do cell death and disruption affect other bodily systems? * Individual level Why does the individual smoke, why can’t the individual stop and what can the individual do to decrease the risk that smoking causes? * Group level Why do certain social groups smoke more than others? * Societal level Why do some societies have higher rates of smoking than others and what can be done legislatively to decrease the rate of harm from smoking? Environmental Interpersonal Individual Genetic Molecular Figure 6.2 Multilevel causation of psychosis 81 Society, place and space Investigations at each of these levels have the same aim – to examine the mechanisms linking smoking to illness. However, the analytical tools needed, and the relevant theory and scientific rules at each level, are different, as are the inferences that can be made from the research. Using the tools and methodology of molecular biology to investigate societal-level factors is unlikely to work very well. Similarly, we are unlikely to understand an individual’s metabolic pathophysiology by using a systems theory of group dynamics. Investigating smoking legislation and tariffs may give an indication of why the rate of smoking is higher in one country than another, it may give information on why there are consequent increases in, for instance, cardiovascular illness in one area or another, but it does not give information on an individual’s risk of harm if he or she is a smoker or why one person smokes and another does not. The possible impact of a factor at any level is constrained by the higher level. To understand causation, we need not only to measure the impact of a factor at a given level, but also to understand the interaction with other levels. There are particular difficulties in studying geographical variations in mental health. In such ecological studies, the exposure (risk factor) is measured at the group level (population or subpopulation) and the outcome is usually the proportion of the group who develop the disease. These aggregated data have some advantages in that there is less random error, more power and less selection bias, but also the disadvantage that one can never be certain that those who developed the disease were those who were exposed. Grouped results tell us about the group but not necessarily about the individuals within the group. For instance, if there was an association between the incidence of schizophrenia and the rate of crime in an area, it could not be assumed that being a victim of crime would increase an individual’s risk of developing the disorder. Making such individual-level inferences from ecological-level data is termed the ‘ecological fallacy’. Further, any systematic differ- ences between areas or groups in the recording or classification of exposure or disease will bias the results in ecological studies. It is also sometimes difficult to identify ecological confounders and effect modifiers, as populations might differ in many ways, such as diet, alcohol consumption and attitudes that may affect the outcome (disease) being measured. Nevertheless, ecological studies are of great value in generating hypotheses and investigating phenomena that only exist at a group level. We now turn to the specific example of urbanicity and psychosis. Urbanisation literature review Incidence rates in rural and urban areas Faris and Dunham (1939) reported that first-admission rates of schizophrenia were particularly high in certain areas of inner-city Chicago, and decreased 82 J. Boydell and K. McKenzie towards the periphery (see Chapter 4 for a detailed summary). Furthermore, there were considerable differences within the inner city itself, with rates being higher in the more disorganised areas, irrespective of deprivation and ethnicity. Faris and Dunham suggested that the nature of living conditions in certain neighbourhoods, particularly extended isolation, produced abnormalities of behaviour and men- tality, which subsequently led to increased rates of schizophrenia. A number of other early studies (reviewed by Freeman (1994)) also demonstrated clearly that the rates of schizophrenia increase towards city centres in Western societies. The generally accepted view for most of the twentieth century was that the high rates of schizophrenia in the inner city could be accounted for by the drift of people with, or beginning to develop, schizophrenia into the more urbanised and anony- mous areas (the ‘social drift’ hypothesis). A variant on this theme suggested non- balanced migration, i.e., that as a city develops, the more able move out to better areas, leaving a residual population in the centre with a high risk of psychiatric disorders (previously called the ‘social residue’ theory). The idea that causal agents are associated with urbanisation (the ‘breeder’ hypothesis) was largely dismissed. It has only been relatively recently that these ideas have been revisited. In a study looking back at the nineteenth century, Fuller Torrey et al. (1997) used the comprehensive 1880 census of the ‘insane’ in the USA to examine the association between urbanicity and severe mental illness. They calculated preva- lence rates for different degrees of urbanicity, characterised as follows: * Urban 30 largest cities; * Semi-urban 50% or more people living in towns of greater than 4000 people; * Semi-rural 25–50% living in towns of 4000 or more people; * Rural 1–25% living in towns of greater than 4000 people; * Completely rural no-one living in a town of greater than 4000 people. Fuller Torrey et al. (1997) found a strong linear trend, with the prevalence in urban areas being 66% higher than in completely rural areas. This study is fascinating, despite its limitations, because a gradient was found between areas that would all be considered rural today. Most recent research has come from Northern Europe where good quality national records have made large-scale epidemiological studies possible. In the first of these, Lewis et al. (1992) investigated the association between place of upbringing and incidence of schizophrenia using data from a cohort of over 49 000 male Swedish conscripts, linking it to the Swedish national psychiatric register. They found a strong significant linear trend. The highest rate of clinically diagnosed schizophrenia was in those who had mostly lived in cities (Stockholm, Go ¨ teborg, Malmo ¨ ) while they were growing up (odds ratio (OR) 1.65). There were intermediate rates in towns with populations greater than 50 000 (OR 1.39), and towns with populations less than 50 000 (OR 1.28), compared with baseline country 83 Society, place and space areas. A similar though weaker trend was found for other psychoses. Adjusting for family finances, parental divorce and family psychiatric history (whether a relative is ‘on medication for nervous trouble’) had little effect on these findings. Adjusting for cannabis use and any psychiatric disorder at conscription reduced but did not eliminate the associations. The authors concluded that causal environmental factors are implicated, as the association remained after adjusting for family history, a proxy measure of genetic risk. However, this study could not distinguish between place of birth, place of upbringing and place of residence at onset. Mortensen et al. (1999) investigated the effect of place of birth on risk of admission with schizophrenia in a large Danish population-based cohort of 1.75 million people. The relative risk of schizophrenia for those born in Copenhagen, compared with those born in rural areas, was 2.40 (95% confidence interval (CI) 2.13–2.7). There was also a clear dose–response relationship for urbanicity: the larger the town of birth, the greater the risk, a finding not explained by a family history of schizophrenia. Mortensen and colleagues further calculated the population-attributable risk (the proportion of the risk of schizophrenia in the whole population that can be accounted for by a particular factor, assuming causality) for urban birth. This was 34.6%, a much larger figure than the 9% and 7% respectively for having a mother or father who suffered from schizophrenia. Like the original study of Faris and Dunham (1939), Mortensen and colleagues found the incidence of manic-depression to be fairly evenly distributed across rural and urban areas. Peen and Decker (1997) also reported a significant positive correlation between admission rates for clinically diagnosed schizophrenia and degree of urbanisation. They also investigated whether differences in the availability of psychiatric services were to blame but concluded that they were not, as the average length of hospital admission and average number of readmissions did not differ between urban and rural areas. There are, of course, problems with using admission data to measure incidence and in relying on national case registers of clinical diagnoses. These strategies enable wide coverage but gloss over the possibility that bias due to different routes to care and diagnostic practices may have affected the results. Such bias can be decreased by studying all incident cases, whether admitted or not, in specified populations. Allardyce et al. (2001) compared all incident cases of psychosis from two areas in the UK, a largely rural part of south-west Scotland and urban south London. The incidence of schizophrenia was 61% higher in the urban area compared with the rural area. Intriguingly, the urban excess was more marked in males than females. The studies of the last decades move us on from considering the urban effect as possibly due to social drift. Dauncey and colleagues (1993), who examined where 84 J. Boydell and K. McKenzie patients had lived five years before their first admission in Nottingham, were unable to find evidence for systematic geographic drift. Though it could be argued that the drift had occurred in previous generations, the magnitude of this movement would need to have been extremely high to explain the findings. For example, in the Danish study, Mortensen and colleagues (1999) calculated that nearly 50 000 children born in the capital city and its suburbs needed to have a parent who transmitted a genetic risk equal to that transmitted by a parent with diagnosed schizophrenia, to account for the urban excess. Social drift in the current or previous generation could also not explain the cumulative effect of urban exposure throughout childhood. This is also relevant for the social residue theory (that those at greater risk are left behind in an area as it becomes less desirable). Krabbendam and van Os (2005) reviewed more recent studies. They report an effect of urbanicity on rates of schizophrenia (i.e., an overall pooled effect size) of 1.72 (95% CI 1.53–1.92), with many showing a dose–response relationship. The effect seems to be greater in younger people and more recent birth cohorts (Haukka et al., 2001; Marcelis et al., 1998) and for schizophrenia. On the basis of their review, Krabbendam and van Os (2005) estimated that urbanicity accounts for around 30% of schizophrenia, assuming causality, in Western countries. Urban birth, upbringing or residence Urbanisation, then, is strongly associated with the incidence of schizophrenia. However, it is unclear whether its effects operate at the time of gestation and birth, throughout the pre-morbid lifespan of the individual, or around the time of onset of the illness. For many years the only published study that had separated place of birth and place of upbringing was that carried out by Astrup and O ¨ degard (1961), which found a stronger effect of city upbringing on the incidence of schizophrenia in those who had moved. More recently, a re-analysis of the Danish population cohort study (Pedersen and Mortensen, 2001) addressed this issue. This study found that schizophrenia risk increased both with the number of years (between 0 and 20 years) that an individual lived in an urbanised area and with increasing degree of urbanisation. Compared with those who had always lived in the most rural areas, the relative risk for those who had spent their entire childhood in the capital was 2.75 (95% CI 2.31–3.28). In contrast, Marcelis et al. (1999) investigated the risk of early onset psychosis in a cohort of all people born between 1972 and 1978. They found no additional risk for urban residence at the time of onset of psychosis above that of urban birth and upbringing. The evidence for an early effect, therefore, is strong, but an effect of urbanisation around the time of onset cannot be completely ruled out. 85 Society, place and space It is now recognised that psychotic symptoms are relatively common in ‘healthy’ people and that there is possibly a continuum between no symptoms and a psychotic illness. Van Os et al. (2001) examined psychiatric symptoms in a large Dutch study of the general population. Psychotic-like symptoms (isolated delu- sions or hallucinations) were found in 17.5% of the community sample, largely in people who had never presented to psychiatric services. Furthermore, such symp- toms increased with increasing urbanisation, but to a lesser extent than for schizophrenia (10–20% higher in urban areas). Possible explanations for the risk-increasing effect of cities Classical causation theory states that we need to demonstrate a correlation, a correct temporal sequence and that no third explanatory variable exists that effects the urban environment and schizophrenia, before concluding that there is a causal link. Most explanations identify the urban environment as either a vector of risk, for instance, by increasing rates of infection, or as a potentiator of known risk factors. But critical analysis demonstrates that the association between urbanicity and schizophrenia is not explained by known risk factors. One problem is that most risk factors have been investigated at the individual level. How plausible, then, are the various proposed explanations? Individual level (1) Biological explanations Genetic risk Although genetic risk itself does not explain the urban–rural difference, as explained above, there is new evidence of gene–environment synergism. Van Os et al. (2004) found that people with a family history of schizophrenia, or any severe mental illness, were more likely to develop schizophrenia if they lived in cities. Obstetric complications Eaton et al. (2000) tested the hypothesis that the urban–rural difference is medi- ated by obstetric complications in a Danish case-control study. They again found that birth in Copenhagen was a strong risk factor for schizophrenia. Obstetric complications that were prospectively measured had a moderate sized relationship to early onset schizophrenia, but the relationship of urbanicity to schizophrenia was unaffected by adjusting for obstetric complications. Similarly, Harrison et al. (2003) found that adjusting for obstetric complications did not significantly reduce the excess incidence of schizophrenia associated with urban birth or upbringing in Sweden. 86 J. Boydell and K. McKenzie [...]... Krabbendam, L and van Os, J (2005) Schizophrenia and urbanicity: a major environmental influence – conditional on genetic risk Schizophrenia Bulletin, 31 (4), 795–9 93 Society, place and space Krieger, N (1994) Epidemiology and the web of causation: has anyone seen the spider? Social Science and Medicine, 39 (7), 887–903 Leighton, A H (1982) Caring for Mentally Ill People: Psychological and Social Barriers... causation should help us understand the importance of a conceptual framework in understanding not only what our findings could mean but also how they can inform intervention Conclusion Investigation of the urban effect demonstrates how difficult it is to understand the impact of society, place and space on the rates of psychosis Even when we try to be more specific and focus on urbanisation, there... 1197–203 McKenzie, K and Harpham, T (2006) Social Capital and Mental Health London: Jessica Kingsley Publishers McKenzie, K., Whitley, R and Weich, S (2002) Social capital and mental illness The British Journal of Psychiatry, 181, 280–3 Mortensen, P., Pedersen, C., Westergaard, T et al (1999) Effects of family history and place and season of birth on the risk of schizophrenia New England Journal of Medicine,... study did not find any interaction between urbanicity and season of birth (Mortensen et al., 1999) or urbanicity and exposure to influenza (Westergaard et al., 1999) Similarly, large population-based studies in Finland (Suvisaari et al., 2000) and the Netherlands (Marcelis et al., 1998) did not find a significant interaction between season and place of birth Overall, it seems that the urban effect... (11), 1039–46 Peen, J and Dekker, J (1997) Admission rates for schizophrenia in the Netherlands: an urban/ rural comparison Acta Psychiatrica Scandinavica, 96 (4), 301–5 Porter, R (1997) The Greatest Benefit to Mankind London: HarperCollins Silver, E., Mulvey, E and Swanson, J (2002) Neighbourhood structural characteristics and mental disorder: Faris and Dunham revisited Social Science and Medicine, 55... W., Mortensen, P and Frydenberg, M (2000) Obstetric factors, urbanization and psychosis Schizophrenia Research, 43, 117–23 Faris, R and Dunham, H (1939) Mental Disorders in Urban Areas Chicago: University of Chicago Press Freeman, H (1994) Schizophrenia and city residence British Journal of Psychiatry, 164 (suppl 23), 39–50 Fuller Torrey, E., Bowler, A and Clark, K (1997) Urban birth and residence as... in Bristol There has been a recent resurgence of 89 Society, place and space interest in this theory Thornicroft et al (1993) noted that clustering of individuals with schizophrenia in deprived areas occurs only in urban areas and suggested that social isolation is an important mediator of this However, it is difficult to distinguish between cause and effect in this context Related to these ideas is... Historical Context New York: Cambridge University Press Lewis, G., David, A., Andreasson, S et al (1992) Schizophrenia and city life Lancet, 340, 137–40 Link, B G and Phelan, J C (1995) Social conditions as fundamental causes of disease Journal of Health and Social Behavior, Extra Issue, 80–94 Link, B G and Phelan, J C (1996) Understanding sociodemographic differences in health – the role of fundamental... et al (1990) Life events and relapse in schizophrenia: a one year prospective study Social Psychiatry and Psychiatric Epidemiology, 25, 221–4 Marcelis, M., Navarro-Mateu, F., Murray, R et al (1998) Urbanization and psychosis: a study of 1942–1978 birth cohorts in the Netherlands Psychological Medicine, 28 (4), 871–9 Marcelis, M., Takei, N and van Os, J (1999) Urbanization and risk for schizophrenia:... area-based research, if arguments over reverse causality are to be overcome The difficulty is that, as psychosis is relatively rare, large areas and numbers of areas are needed to gather sufficient cases, and that, together with the 91 Society, place and space intensive interviewing of the cases needed to determine the presence of individuallevel risk factors, makes a definitive study extremely expensive . 6 Society, place and space Jane Boydell and Kwame McKenzie Introduction This chapter will discuss the impact of society, place and space on the. impact on the group (McKenzie and Harpham, 2006, pp. 86–109). Conceptualising society, place and space Over time, society, place and space have been conceptualised