Migration, ethnicity and psychosis

10
Migration, ethnicity and psychosis
Kwame McKenzie, Paul Fearon and Gerard Hutchinson
Introduction
In the 1930s, O
¨
degaard (1932) reported that first-admission rates for schizo-
phrenia were high among Norwegian migrants to the United States. Since then
numerous studies in a variety of countries have investigated rates of serious mental
illness in migrant groups and in different cultural and ethnic groups within
countries (Cantor-Graae and Selten, 2005). In this chapter, we review the literature
reporting differences in the incidence of psychosis between migrant and ethnic
groups, we discuss methodological issues and, using the best-researched group,
people of African and Caribbean origin in the UK, we try to build a model of how
migration, culture and ethnicity affect rates of incident psychosis.
History and overview
Since high rates of mental illness among Norwegian migrants to the United States
were reported in the first half of the last century (O
¨
degaard, 1932) there have
been a number of studies investigating the incidence of psychosis in migrant and
ethnic minority groups. Although the vast majority of migration is between
developing countries within Africa and Asia, there is surprisingly little research
on the risk of psychosis in these groups. Research into the incidence of psychosis
in migrant groups is best developed in northern Europe. The most comprehensive
literature on the subject concerns the high incidence of schizophrenia in people
of African and Caribbean origin who migrated to the UK, mainly in the 1940s
and 1950s, and in their children and grandchildren, a finding which has been
consistently reported for 30 years. The various studies have reported rates of
schizophrenia between 2 and 14 times greater for African-Caribbeans than for
whites in the UK (Fearon and Morgan, 2006). Some of the more recent findings

are shown in Table 10.1. Studies have also reported elevated rates in migrant
compared with host populations in other northern European countries, including
Society and Psychosis, ed. Craig Morgan, Kwame McKenzie and Paul Fearon. Published by Cambridge
University Press. # Cambridge University Press 2008.
the Netherlands (Selten and Sijben, 1994; Selten et al., 1997; Selten et al., 2001),
Denmark (Cantor-Graae et al., 2003) and Sweden (Zolkowska et al., 2001) and in
Australia (Krupinski and Cochrane, 1980). Cantor-Graae’s and Selten’s (2005)
meta-analysis demonstrated a significant increased risk of schizophrenia in all
migrant groups, this being greatest in those from developing countries who
migrated to developed countries and in those with black skin colour migrating
to countries where the population was predominantly white.
There has been much speculation about why rates of psychosis are raised in
migrant and ethnic minority groups. O
¨
degaard (1932) suggested that the increase
might be due to selection, i.e., those with a predisposition to psychotic illness being
more likely to migrate, but recent studies have refuted this hypothesis (Selten et al.,
2002). It seems unlikely that biological factors can explain the high rates, and
because of this the focus has shifted to social experiences and conditions
(Sharpley et al., 2001). Over time, the reasons for migration, the processes of
migration and the situations of migrants in their new host countries have all come
under scrutiny as researchers have sought to explain the high rates (Sharpley et al.,
Table 10.1 Reported incidence rates for narrowly defined schizophrenia
(per 10 000 per year) in African-Caribbeans in recent UK studies and in Barbados,
Trinidad and Jamaica
Incidence rate per
10 000 per year
a
95% confidence
interval

UK 7.1 (5.2–9.0)
(Fearon et al., 2006)
UK 5.1 Not given
(Bhugra et al., 1997)
UK 4.7 (1.8–7.5)
(Harrison et al., 1997)
UK 5.3 (1.8–8.7)
(King et al., 1994)
Barbados 2.8 (2.0–3.7)
(Mahy et al., 1999)
Trinidad 1.6 (1.1–2.1)
(Bhugra et al., 1996)
Jamaica 2.1 Not given
(Hickling and Rodgers-Johnson, 1995)
a
The WHO Ten Country study found rates of narrowly defined schizophrenia that ranged from
0.7 to 1.4 per 10 000 per year (Jablensky et al., 1992). The meta-analysis by McGrath et al. (2004)
found a median incidence of 1.5 per 10 000, based on 100 incidence studies.
144 K. McKenzie, P. Fearon and G. Hutchinson
2001). As migrants have had children, they have also become the focus of inves-
tigation (Bhugra et al., 1997; Castle et al., 1991; Fearon et al., 2006; Harrison et al.,
1988; Harrison et al., 1997; McGovern and Cope, 1987; Thomas et al., 1993; van
Os et al., 1996). This, and the possibility that a mismatch between the culture of
migrant groups and the host population is aetiologically important, has led to
a link between the study of migration and ethnicity (see definitions below). This
said, there remain a number of methodological problems with studies of the
incidence of psychosis in migrant and ethnic minority groups, and these need to
be considered in drawing conclusions. Before doing this, it is necessary to define
some relevant concepts and terms.
Definitions of relevant concepts and terms

Migration
Migration can be considered as a process of social change in which an individual
moves from one cultural setting to another. There are many reasons for, and types
of, migration, these often being enshrined in law in host countries (e.g., temporary
workers, economic migrants, asylum seekers and refugees). A further distinction
can be drawn between primary migrants and secondary migrants who follow in
their footsteps. Three stages to the migration process have been identified (Bhugra
and Jones, 2001):
*
Pre-migration decision to move and planning;
*
Transition movement from one setting to another;
*
Post-migration coming to terms with a new life, roles and country.
The factors that increase risk for disorder in migrant groups may operate during
any of these stages.
Race, culture and ethnicity
Race, culture and ethnicity are three related but distinct concepts. The idea that
people can be separated into racial categories on the basis of physical appearance
has a long history in the West, and the popular belief that people are separable into
distinct groups on the basis of phenotypical characteristics persists and underpins
ongoing racism (Fernando, 1991). Modern genetics has undermined the scientific
validity of racial categories. For example, it has been shown that the differences
between classically described racial groups (10% of the genetic variation) are only
slightly greater than those which exist between nations (6%), and both of these are
small compared with genetic differences within local populations (84%) (Jones,
1981). The use of racial categories has now largely disappeared from scientific
research and been replaced by the use of ethnicity. However, researchers often
145 Migration, ethnicity and psychosis
categorise people into ‘ethnic’ groups in such a way that these are indistinguishable

from racial categories, e.g., the crude dichotomy between black and white (for a
review see McKenzie and Crowcroft, 1994).
There are many definitions of culture, but what is common to most is the idea
that culture provides a set of socially shared guidelines or rules that shape and
constrain beliefs, attitudes and behaviour. In other words, culture usually refers to
the behaviours and attitudes of social groups. That said, culture is not static or
homogeneous. Determined by upbringing and choice, culture is constantly chang-
ing and is notoriously difficult to measure (Fernando, 1991). Such cultural flux
may be particularly important in groups who have migrated, where individuals
face choices about how much of the host culture to incorporate into their own.
Of all the variables, ethnicity is probably the most difficult to define and use.
Such groups are characterised by a sense of belonging or group identity (Jenkins,
1986), these being dynamic and changeable and determined by social pressures
and psychological needs. Aspects of race and culture may engender a sense of
common identity and so, in part, determine ethnicity. For example, as Fernando
(1991) suggests, a sense of belonging may emerge from the shared experiences of
discrimination in a racist society – emergent ethnicity. This may be one factor that
has driven the emergence of a Caribbean identity among migrants from the
culturally diverse Caribbean islands to the UK. Cultural heritage may form a
significant component of ethnic identity, but it does not define it, and those
who perceive themselves as belonging to an ethnic group may well differ markedly
in terms of the cultural reference points that inform their beliefs and actions. This
warns against the conflation of culture and ethnicity.
Ethnicity is potentially fluid and changeable over time and space, as exposure
to other contexts and cultures allows for its reformulation. However, the way in
which ethnicity is measured and operationalised in much epidemiological research
ignores these complexities. The use of fixed, predetermined ‘ethnic categories’ in
cross-sectional research and by governments for census purposes is problematic in
that key components of ethnicity – sense of belonging and changeability – are
absent. For example, in one US study, in which subjects were asked to select their

ethnicity at baseline and 12 months later, one third selected a different ethnicity at
the second time of asking (Leech, 1989).
Methodological problems
For all the apparent consistency of studies showing high rates of psychosis in
migrant groups (Cantor-Graae and Selten, 2005), there are a number of methodo-
logical issues that need to be considered in evaluating the validity of the findings.
What we are ultimately trying to do is model and understand complex social
146 K. McKenzie, P. Fearon and G. Hutchinson
processes with multiple layers and meanings. To do this effectively, methodo-
logical rigour is essential, as is caution in interpreting results.
Measuring migration, ethnicity and culture
Migration is a complex variable but many studies simplify it to being a member of
a migrant group or not. Incidence studies are unable to indicate whether members
of the groups are economic migrants, refugees or asylum seekers, as they rely on
census estimates and categories, which do not distinguish these groups, for their
denominator. Perhaps the single greatest difficulty is the measurement of context.
Not only do different groups have different experiences of the first two stages of
migration (see above), but they also migrate into different sociopolitical contexts.
These contexts are likely to be important in more fully understanding the risks
for psychosis to which migrants may be exposed. A migrant group may include
different cultural or ethnic groups. If information describing the process of
migration and the different ethnicities and cultures within a migrant group is
lacking, it becomes difficult to understand precisely what the risk-generating
exposures are.
Quantifying ethnic groups
Accurate enumeration of ethnic minority populations is an important, but prob-
lematic, issue. For example, national censuses, which are often used to provide
the denominator for calculations of incidence rates, have varying levels of accuracy
for different ethnic groups. Many national censuses give only estimations of ethnic
minority populations and the size of the error can vary significantly by ethnic

minority group. For example, the 1991 UK census underestimated the population
by approximately 1 million people, and those not counted in this ‘missing million’
were disproportionately young, male and from ethnic minority groups. Further-
more, in the UK census, respondents self-assign ethnicity to predetermined
groups. However, in some studies, the ethnicity of incident cases is observer-
assigned (e.g., Harrison et al., 1997). Self-assigned and observer-assigned ethnicity
may capture different people, so when studies use the latter, their numerator
(number of cases of psychosis per ethnic group) is actually measuring a different
group from their denominator (number of people in the defined population per
ethnic group).
Ascertainment bias
Most incidence studies of psychosis measure first-admission or first-contact cases.
This produces a number of challenges. For example, people from different ethnic
and cultural groups with the same diagnosis follow different pathways to care, and
the risk of an inpatient stay may vary (for a review see van Os and McKenzie,
147 Migration, ethnicity and psychosis
2001). This may undermine the accuracy of estimates of incidence based on first
admission (van Os and McKenzie, 2001). First contact studies, which include not
only first-admission but also first-presentation cases to primary care and com-
munity mental health services, have been proffered as more accurate. However,
there are no studies that allow estimation of any likely error rate one way or the
other. In other words, incidence rates based on treated samples of cases may differ
from the ‘true’ population incidence; if people from ethnic minority groups are
more likely to come into contact with health services when they develop psychotic
symptoms this could artificially increase the ‘incidence’ in these groups. There are
no good data that allow us to estimate the consequent error.
Validity testing and the category fallacy
There is no doubt that psychosis is present across the world. However, quantifiable
differences in the aetiology and course of psychosis in ethnic minority groups have
been reported, and this brings into question whether, in cross-cultural incidence

studies, like is being compared with like (Harrison et al., 1999; McKenzie et al.,
2001). The cross-cultural validity of current categorical diagnoses, and the data
collection tools based on them, has not been formally tested (Alarc
´
on et al., 2002).
Though it is clear that well defined core symptoms of schizophrenia can be found
everywhere in the world, if we argue that a particular aetiological insult, e.g.,
migration, could lead to an increased risk of psychosis, this may equally point to
the possibility of new aetiologies or quantifiable differences in the balance of risk
factors between groups. These could produce different rates of the same illness or
new forms of the illness. Hence, the tools we use must not only be cross culturally
valid, but must also be able to detect and differentiate new or atypical psychoses
produced by new risk factors.
A critical review of the findings
The methodological problems should rightly make us cautious in interpreting
cross-cultural incidence studies, but they do not negate them. Our aim has been to
identify possible limitations so that we can be judicious in the way these findings
are used to generate further hypotheses. So, turning to the research in more detail,
what conclusions can be drawn? In addressing this question, as the best data are on
schizophrenia, we will use this to illustrate the challenges posed by the current
literature.
Migration and schizophrenia: a meta-analysis
Cantor-Graae and Selten (2005) recently conducted a meta-analysis of studies
of schizophrenia and migration. Their criteria for inclusion were that studies:
148 K. McKenzie, P. Fearon and G. Hutchinson
(1) reported schizophrenia incidence rates for one or more migrant groups
residing in a particular area or provided data so that such a calculation could be
performed; (2) included a correction for age differences between groups or
provided data that made this correction possible; and (3) were published in an
English-language, peer-reviewed scientific journal. When there were two studies

on the same ethnic minority groups in the same area in the same time period, only
one study was chosen.
Eighteen studies were included. Of these, eight were first-contact studies
(Bhugra et al., 1997; Castle et al., 1991; Goater et al., 1999; Harrison et al., 1988;
Harrison et al., 1997; Rwegellera, 1977; Selten et al., 2001; Zolkowska et al., 2001)
and ten were first-admission studies (Cantor-Graae et al., 2003; Cochrane and Bal,
1987; Dean et al., 1981; Hitch and Clegg, 1980; Krupinski and Cochrane, 1980;
McGovern and Cope, 1987; Selten and Sijben, 1994; Selten et al., 1997; Thomas
et al., 1993; van Os et al., 1996). Only five studies included more than 100 patients
from migrant groups (Cantor-Graae et al., 2003; Cochrane and Bal, 1987; Dean
et al., 1981; Krupinski and Cochrane, 1980; Selten et al., 1997). Studies often
combined ethnic groups (as defined above) and groups categorised by place of
birth. Only one study was conducted outside Europe and this investigated
European-born migrants to Australia (Krupinski and Cochrane, 1980). Two-
thirds of the studies were from the UK (Bhugra et al., 1997; Castle et al., 1991;
Cochrane and Bal, 1987; Dean et al., 1981; Goater et al., 1999; Harrison et al., 1988;
Harrison et al., 1997; Hitch and Clegg, 1980; McGovern and Cope, 1987;
Rwegellera, 1977; Thomas et al., 1993; van Os et al., 1996).
The most common group studied was people of Caribbean origin in the UK
(Bhugra et al., 1997; Castle et al., 1991; Cochrane and Bal, 1987; Dean et al., 1981;
Harrison et al., 1988; Harrison et al., 1997; Hitch and Clegg, 1980; McGovern and
Cope, 1987; Rwegellera, 1977; Thomas et al., 1993; van Os et al., 1996); the next
most studied group was people of South Asian origin in the UK (Bhugra et al.,
1997; Cochrane and Bal 1987; Dean et al., 1981; Goater et al., 1999; Hitch and
Clegg, 1980; Thomas et al., 1993). These groups are a mixture of migrants, their
children and their grandchildren, although the authors did distinguish between
first and subsequent generations where possible. First-contact replication studies
are reported for people of Caribbean origin, but only first-admission replication
studies are reported for people of South Asian origin. Only four other groups were
the focus of more than one incidence study, i.e., people who migrated to the

Netherlands from Surinam, Dutch Antilles, Morocco and Turkey.
The meta-analysis of these studies produced a mean weighted relative risk for
developing schizophrenia among first-generation migrants (40 estimates) of 2.7
(95% CI 2.3–3.2). In a separate analysis for second-generation migrants (7 estimates)
the mean weighted relative risk was 4.5 (95% CI 1.5–13.1). However, this was based
149 Migration, ethnicity and psychosis