Incidence of psychotic illness in London: comparison of ethnic groupsBMJ 1994; 309 doi: http://dx.doi.org/10.1136/bmj.309.6962.1115 (Published 29 October 1994) Cite this as: BMJ 1994;309:1115
- M King,
- E Coker,
- G Leavey,
- A Hoare,
- E Johnson-Sabine
- University Department of Psychiatry
- Royal Free Hospital School of Medicine
- London NW3 2QG Department of Psychiatry
- St Ann's Hospital
- London N15 3TH
- Correspondence to: Dr King.
- Accepted 19 August 1994
Objective: To compare annual incidences of20psychosis in people from different ethnic groups as defined in the 1991 census.
Setting: Catchment area of district psychiatric20hospital. Design - All people aged 16 to 54 years who made contact with a wide range of community and hospital services between 1 July 1991 and 30 June 1992 were20screened for psychotic symptoms. Patients with such symptoms were interviewed face to face to collect information on demography, ethnic group,20psychiatric history and symptoms, drug use, and how care had been sought. A key informant, usually a close relative, was also interviewed.
Main outcome measures: Age standardised incidence of schizophrenia and non-affective psychosis according to the ninth edition of the International Classification of Diseases in each ethnic group. Results - Ninety three patients took part, of whom 38 were assigned a certain or very likely diagnosis of schizophrenia (15 in white population, 14 in black,20seven in Asian, and two in others). The age standardised annual incidence of schizophrenia was 2.2 (95% confidence interval 1.5 to 2.9) per 10 000 of the population. The incidence ratio for schizophrenia in all ethnic minority groups compared with the white population was 3.6 (1.9 to 7.1); the corresponding figure for non- affective psychosis was 3.7 (2.2 to 6.2).
Conclusions: Raised incidences of schizophrenia were not specific to the African Caribbeans, which suggests that the current focus on schizophrenia in20this population is misleading. Members of all ethnic minority groups were more likely to develop a psychosis but not necessarily schizophrenia. The personal and social pressures of belonging to any ethnic minority group in Britain are important determinants in the excess of psychotic disorders found.
Schizophrenia is reported to be more prevalent among African Caribbeans living in Britain than among other ethnic minority groups and the white population
This study of the annual incidence of psychosis among different ethnic groups (defined according to the 1991 census) found that the incidence of schizophrenia was higher among all ethnic minority groups, not just black Caribbeans
Ethnic group bore no relation to the mode of onset or clinical picture of a first episode of psychosis or the likelihood of being admitted
Cannabis use was not associated with any particular ethnic group or with then onset of psychosis
The current focus on schizophrenia in African Caribbeans is misleading as all ethnic minority groups are vulnerable
The variation in the incidence and prevalence of schizophrenia in different ethnic groups in the United Kingdom remains controversial. Although there is disagreement about admission rates to psychiatric hospital for Asians with psychosis,1 African Caribbeans have higher admission rates and receive a diagnosis of schizophrenia more often than do members of other ethnic groups.*RF 2-4* Possible explanations include selective migration,5 a negative experience after migration and the influence of racism,6 racial prejudice in psychiatric practice,7 inequitable service utilisation,8 use of cannabis,3 and specific biological causes.9 There has been considerable criticism of the findings, particularly with regard to the possibility of diagnostic inaccuracy,6 errors in calculating denominators,10 and a lack of appropriate comparison groups.6
To our knowledge the only other prospective study in Britain that focused on ethnic group was by Harrison et al.11 They found that the mean annual incidence of schizophrenia in African Caribbeans aged 15-54 years was 12 times higher than that in the remaining population and 18 times higher than that in those who were British born and aged 16-29 years. They, however, did not consider other ethnic groups or use a concurrent comparison group, and they used indirect figures to derive the population denominator.
We assessed the annual incidences and rate ratios of psychotic mental disorder in people from all ethnic groups resident in the defined catchment area of a psychiatric hospital in north London in the census year 1991; 1991 was the first census to estimate ethnic group directly. The black community in Haringey is one of the largest in a single health district in England and Wales.12
Patients and methods
Entry to the study
All people aged 16 to 54 years who were resident in the catchment area of St Ann's Hospital and made contact with health services between 1 July 1991 and 30 June 1992 were eligible to take part. Information on potential cases was obtained from community, domiciliary, outpatient, casualty, and inpatient services. The community services included all local general practitioners, a mental health clinic, community psychiatric nurse case managers, and the community psychiatric team. All 280 general practitioners in the 125 practices of the Enfield and Haringey Family Health Services Authority were circulated with a letter (which included the screening criteria) every month for the year of recruitment. Members of staff of the community mental health teams and the regional interim secure unit and medical staff in Brixton and Pentonville prisons, the main remand prisons for the population, were contacted monthly by post. Medical staff in all inpatient facilities were contacted monthly by post and visited personally every two weeks by a member of the research team. All the professionals concerned were advised that we were seeking all patients with new onset of psychotic illness, but our specific hypotheses were not discussed. Patients from the catchment area who were admitted outside the area were monitored by a search of the integrated district and regional information system for the North East Thames Health Region.
We screened the information on potential cases for possible features of psychotic disturbance. The screening criteria applied were those used by Harrison et al.11 In summary, evidence was required for at least one of the following: hallucinations, delusions, obvious thought disorder, psychomotor disorder, bizarre or grossly abnormal behaviour, and a definite change in personality or behaviour. Exclusion criteria were acute or chronic organic brain syndromes and conditions attributable to alcohol abuse. Patients meeting these criteria were approached to take part in a detailed, structured, face to face assessment which was carried out by two of us (EC and AH). EC is of black African origin born in England and AH is white and also born in England.
The study was approved by all relevant ethical practices committees. Each patient gave informed consent; consent was also obtained from the next of kin of patients detained under a section of the Mental Health Act 1983.
Face to face interview was used to collect demographic details, a conventional psychiatric history using a semistructured questionnaire, an assessment of psychiatric symptoms using the present state examination,13 and an assessment of how the patient had sought help. A key informant, usually a close relative, was interviewed to corroborate the history and provide details of onset of disturbance and how care had been sought.
Drug use - Each subject was asked standardised questions about the use of non-prescribed drugs and was asked to give a urine sample, which was screened for drugs.
Ethnic group was assigned in the same way as the 1991 census. When patients were unable to assign themselves an ethnic category the interviewer assigned a category following the guidelines laid down by the Office of Population Censuses and Surveys. The ethnic ratings used in the 1991 census were arrived at after extensive consultation with ethnic minority organisations and piloting for acceptability during the 1980s.14 It provides the most accurate measure yet of the ethnic composition of the British population.
Psychiatric diagnosis was assigned to each patient according to the ninth edition of the International Classification of Diseases (ICD-9)15 and the revised third edition of the Diagnostic and Statistical Mannual of Mental Disorders (DSM-III-R) of the American Psychiatric Association.16 When there was uncertainty, an alternative diagnosis was also recorded. Each diagnosis was the consensus of discussion between at least three clinical members of our research team (MK, EC, EJ-S, and AH). Only the original interviewer (EC or AH) was aware of the subject's ethnic origin. On the basis of the ICD consensus diagnosis, the degree of certainty for the diagnosis of schizophrenia was established as described previously.11,17
To estimate incidence a broad classification of non-affective psychosis was derived; this included schizophrenia (ICD-9 code 295) and paranoid states (ICD-9 code 297) and other non-organic psychosis (ICD-9 code 298). ICD- 9 codes 297 and 298 were, however, also examined in their own right. The CATEGO class, a computer estimate derived from the present state examination, was used to measure symptoms in the previous month.18
Onset of disorder - The mode of onset of disturbance was established from information given by the patient and key information at interview and from case notes. As much independent information as possible was used to assess time since the first detectable change in personality and behaviour or the development of unequivocal psychotic symptoms, or both. Onset was further categorised as acute or insidious according to the classification established by the World Health Organisation.19
Univariate analyses applied were the X2 statistic for proportions, the Mantel-Haenszel test for linear association, and Student's t test and analysis of variance for differences in means. Incidences were calculated from the 1991 census figures for the catchment area, which were made available to us by the health authority. Rates were standardised in two broad age bands (16 to 29 and 30 to 54 years) using figures for England supplied by the Office of Population Censuses and Surveys. Confidence intevals for incidences were calculated with the computer program confidence intervals analysis,20 and rates were compared by linear logistic modelling with the GLIM computer program.21
One hundred and three subjects satisfied the screening criteria for having a possible first onset psychotic disorder. Three patients refused to take part and seven could not be contacted. A search of the integrated district and regional information system found a further three psychotic patients who had been missed. Thus, 93 patients (88%) were successfully recruited, five of whom were unable to undergo a present state examination owing to disturbed behaviour and concentration. An ethnic classification was attempted on the 10 patients who satisfied criteria but refused to participate or could not be contacted. Four were white, three black Caribbean, two black African, and one black other.
The mean ages were 31.9 years and 27.2 years for women and men respectively (mean difference 4.7 years (95% confidence interval 1.1 to 8.4), t=2.53, df=91, P=0.013) (table I), The median time of residence in the area was 39 months, 14 patients having lived there for six months or less. Ten (11%) patients had refugee status compared with 5% of people (of all ages) in the population of the catchment area.22 Fifty four patients were born in the United Kingdom; in 40 cases one or both parents had been born abroad, in 11 both parents had been born in the United Kingdom, and in three the parents' place of birth was uncertain.
Sixteen patients were unwilling or unable to decide on their ethnic group, which was therefore assigned by the interviewer (table I). Census figures for the St Ann's catchment area are shown in table II. Of the 39 white patients, 18 were of British origin, five were Irish, seven were Greek or Turkish Cypriot, and nine were of other European origin.
There were no significant differences between the ethnic groups as categorised in the 1991 census (white, black Caribbean (African Caribbean), black African, black other, Asian (Indian, Pakistani, and other Asian) and other) in terms of sex, marital state, social class of head of household, level of education, employment, period of residence in the area, or homelessness. White patients were older than the others (mean difference 3.4 years, P=0.08).
Incidences of psychotic illness
Sixty two patients were given a diagnosis in one of the categories of non- affective psychosis, including schizophrenia (ICD-9 codes 295, 297, and 298). Thirty eight patients were assigned a restrictive (certain or very likely) diagnosis of schizophrenia (table III). There was strong agreement between ICD-9 and DSM-III-R; only one of these 38 patients received a discrepant diagnosis. When the Mantel-Haenszel test for trend was applied there was no significant relation between the degree of certainty of schizophrenia (five categories in table III) and ethnic group.
The annual incidences of schizophrenia and non-affective psychosis were standardised for age with a cut off point of 29 years for ease of comparison with other studies in which incidences have not been standardised (table IV).11 The age standardised annual incidence of schizophrenia was 2.2 per 10 000 population (1.5 to 2.9). Linear logistic modelling (controlling for age) showed significant overall differences between the white population, black Caribbeans, black Africans and black others, all Asians and others in the rate of schizophrenia (X2=19.54, df=4, P=<0.001) and non-affective psychosis (X2=30.51, df=4, P<0.001). To avoid multiple comparisons of small numbers, we report rate ratios only for the white population against all other ethnic groups. The incidence ratio for schizophrenia in all ethnic minority groups compared with the white population was 3.6 (1.9 to 7.1); the corresponding ratio for non- affective psychosis was 3.7 (2.2 to 6.2).
Nine patients (five white, two black African, one black other, and one other) three of whom were refugees, received a diagnosis of mania (ICD-9 codes 296.0, 296.2, and 296.4). The age standardised annual incidence of mania in the population of the catchment area was 0.5 (0.2 to 0.8) per 10 000 population.
Nine of the 10 refugees (five white, four black African, and one other) were given a diagnosis of affective psychosis (eight of ICD-9 code 296 and one of ICD-9 code 298). One was given a diagnosis of non-affective psychosis (ICD-9 code 297.8, paranoid state); in this case there was a degree of uncertainty about the diagnosis.
Doubtful interviews, in which the adequacy of ratings of key symptoms in the present state examination was questionable, occurred no more or less often in any particular ethnic group. The ethnic group of the interviewer was not a factor influencing the rate of doubtful interviews. On the basis of the CATEGO class there were no unusual or distinctive syndromes predominating in any particular ethnic group.
The distribution of the time that had elapsed between the first sign of a change in behaviour or personality and the first contact with psychiatric services was highly skewed, with a mean of 90 weeks and a median of 28 weeks, but it did not vary significantly between the white, black, Asian, or other ethnic groups. In particular, there was no difference between black Caribbean and other patients in time since the first onset of schizophrenia or all psychoses (median difference 2 weeks longer for black Caribbeans, P=0.33). There was no difference between the ethnic groups with regard to the type of onset of the disorder, broadly defined as acute or insidious.
Fourteen patients had a first degree relative who had had a psychotic illness. Neither size of sibship nor prevalence of psychotic illness in first degree relatives differed significantly between ethnic groups.
Black patients were overrepresented among those with less common psychotic syndromes: five black Caribbean, three black African, and two white patients had diagnoses with ICD-9 codes 295.4, 298.3, 298.4, 298.8, and 298.9). Only five of these, however, had developed the disorder acutely (in two weeks or less).
All nine black Caribbeans with schizophrenia were aged 29 years or under, whereas there was a broader age distribution in the other ethnic groups.
More white patients (33/39, 85%) had been admitted to hospital at ascertainment compared with black patients (23/38, 61%). Asians (7/11, 64%), or patients in the census classification other groups (2/5, 40%) (X2=7.93, df=3, P=0.047). There were no differences between the ethnic groups in terms of compulsory admission under the Mental Health Act or the police being involved in admission.
Use of non-prescribed drugs
Cannabis was the main drug misused. Fourteen (36%) white, 12 (32%) black, and three (27%) Asian patients reported having used cannabis at one time. Six white and seven black patients used cannabis at least weekly. Urine samples were screened for drugs in 63 patients; six tests were positive for cannabinoids, four of which were from black patients. There was no association between ICD-9 diagnosis of schizophrenia and use of cannabis.
To our knowledge, ours is the only prospective study in Britain of first onset psychosis to have focused on all ethnic groups in a diagnostic interview. The age standardised annual incidence of schizophrenia was 2.2 (1.5 to 2.9) per 10 000 population, close to the rate reported previously in a British study.17 The annual incidence of mania was just below the lower range usually quoted for mania in Britain.23 Annual incidences of schizophrenia or all non-affective psychoses were consistently higher in the ethnic minority groups than in the white population. Confidence intervals for the estimated incidences were wide for some groups, particularly the Pakistanis, because of their small numbers; therefore these figures must be regarded with caution. Littlewood and Lipsedge pointed out that African Caribbeans present with schizophrenia at a younger age.6 This is confirmed by our results (schizophrenia in African Caribbeans occurred only in patients aged 16 to 29 years) and supports their view that lifetime incidence in African Caribbeans may be closer to the incidence in the base population.
The incidence of schizophrenia in African Caribbeans (all of whom were second generation) relative to other groups was lower than that reported by Harrison et al in Nottingham.11 The 1991 census enabled us to obtain a much more accurate estimate of the denominator population and avoided the need for extrapolations from other sources, as in the Nottingham study.11 As far as possible, we also limited the confounding influence of use of hospital psychiatric services by recruiting patients from primary care and community sources as well as from hospital psychiatric services.23
Despite the higher rates, we found little in the mode of onset or clinical picture, which distinguished the ethnic groups. Demographic factors were not important confounders. Service factors, such as admission, did not distinguish the ethnic groups, and we could not confirm other work that police involvement or use of the Mental Health Act were more likely in black patients.3,24 Reported cannabis use was not associated with any particular ethnic group and was thus unlikely to be an explanation for the higher incidence of psychosis in ethnic minorities.
First and second generation migrants
Like others,11,25 we found little evidence for the “atypical psychosis” - characterised by a female preponderance, a sudden onset, prominent emotional symptoms, and florid (often religious) delusions - that has been reported in African Caribbean people living in Britain, particularly first generation migrants.4 We did, however, find more unusual psychotic syndromes in black patients. Possibly because we assessed time of onset in a standardised manner, which is often not the case in routine clinical work, we found few truly acute syndromes and considerable concordance between the criteria of the ICD-9 and DSM-III- R; the criteria of DSM-III-R require at least six months of symptoms to diagnose schizophrenia.
Our population was striking in that 85% were born abroad or had at least one parent who had been born abroad. Most white patients with schizophrenia were also first or second generation migrants. Although some have speculated that raised incidence in African Caribbeans may have a biological basis,9 the consistent finding over many years and in several countries that first or second generation migrants of any ethnic group may be at greater risk of a psychotic illness is suggestive of environmental rather than constitutional factors in the provocation of these disorders. The sources of stress on members of ethnic minorities include the effects of migration, discrimination, racism, and cultural change. Perhaps the most important determinant of the mental health of ethnic minorities in Western countries is the conditions under which they live.26 However, why such stresses should result in psychotic disorder rather than other more usual expressions of stress, such as anxiety disorders and alcohol dependence, is unclear. Nine of the 10 refugees developed an affective psychosis. We cannot say whether these disorders are provoked by the stress of this particular form of migration, but they stand in obvious contrast with the other psychoses occurring in people migrating for other reasons. These issues are best examined by case-control studies, in which index patients are compared directly with random controls selected from the same catchment area. We are undertaking such a study to try to unravel some of these important points.
Hospital admissions for psychotic disorders in people from the Indian subcontinent have been reported as being both higher and lower than those in the base population14,25; differences exist within the Asian community, Pakistani born women having significantly lower rates.27 The relatively small Asian population living in the catchment area of St Ann's Hospital may reduce the role of the extended family and community support that is postulated to be an important protective factor against severe mental illness in this community.28 Sound evidence for this cultural stereotype, however, is lacking.26 Our method of ascertainment may have avoided selection factors, such as differential use of services,29 thereby estimating incidence more accurately. Only a similar study with larger numbers will confirm this finding.
Limitations of study
There are important limitations to our results. The census figures may not be completely accurate; between 0.57% and 2% of the population may have avoided the census.30 The Office of Population Censuses and Surveys, however, has recently reported on a comparison between the 1991 census figures and the data of the 1989-91 labour force surveys, concluding that there is unlikely to be an ethnic bias in the rates of underenumeration.14 The census figures for the area indicated that black Caribbean men may be underrepresented by up to 10%. Assuming underenumeration by all black Caribbeans to this degree reduces their standardised incidence of schizophrenia only from 5.3 to 4.8 (1.7 to 7.9) per 10 000 population. A further limitation is that diagnosis in first onset psychosis is not always straightforward. Patients may display symptoms or behaviour which do not fit a diagnostic category because the symptoms have not been present for long enough. The presentation of mental illness may differ in different cultures and ethnic groups,4 and the use of categorical diagnoses established in Western populations may miss important cultural distinctions in the presentation of psychological distress.31 The CATEGO class derived from the present state examination, however, did not confirm the presence of unusual or distinctive syndromes in any particular ethnic group. We assessed symptoms cross sectionally at first presentation; the predictive validity of orthodox, psychiatric diagnosis can be established only by determining outcome. All patients recruited to the study will be interviewed again after 12 months.
In conclusion, we regard the current focus on African Caribbeans and schizophrenia as misleading. Members of all ethnic minority groups are more likely to develop a psychosis but not necessarily schizophrenia. Although there is a biological component to psychosis, this finding suggests that the personal and social pressures of belonging to any ethnic minority group in Britain are important determinants of the excess of psychotic disorders found.
This study was supported by the Department of Health and the North East Thames Regional Health Authority. We thank all the patients and their relatives who took part and the clinical and administrative staff in the Haringey health district for their helpful cooperation throughout the study.