Ethnic variations in pathways to and use of specialist mental health services in the UK

Abstract

Background

Inequalities of service use across ethnic groups are important to policy makers, service providers and service users.

Aims

To identify ethnic variations in pathways to specialist mental health care, continuity of contact, voluntary and compulsory psychiatric in-patient admissions; to assess the methodological strength of the findings.

Method

A systematic review of all quantitative studies comparing use of mental health services by more than one ethnic group in the UK. Narrative analysis supplemented by meta-analysis, where appropriate.

Results

Most studies compared Black and White patients, finding higher rates of in-patient admission among Black patients. The pooled odds ratio for compulsory admission, Black patients compared with White patients, was 4.31 (95% CI 3.33–5.58). Black patients had more complex pathways to specialist care, with some evidence of ethnic variations in primary care assessments.

Conclusions

There is strong evidence of variation between ethnic groups for voluntary and compulsory admissions, and some evidence of variation in pathways to specialist care.

The Human Rights Act 1998 and the Race Relations (Amendment) Act 2000 place responsibilities on service providers to eradicate discriminatory procedures and practice. Inequitable variation in use of health care services between ethnic groups concerns policy makers, patients and clinicians (Acheson, 1998; Department of Health, 2000). A previous review reported that mental health services are unattractive to some ethnic groups, who complain of more coercive treatments and adverse experiences (Cochrane & Sashidharan, 1996). Although this perception has become widespread, studies that compare access and use of mental health services by different ethnic groups have not been reviewed systematically. It is not known how consistently the findings suggest poorer access for minority ethnic groups, or how these findings relate to study design and methodological quality (Cochrane & Sashidharan, 1996). We undertook a review of the published studies on ethnic variations in use of primary care or mental health services in the UK.

METHOD

Our review focused on specific questions: is there variation between ethnic groups in (a) pathways to specialist mental health care; (b) continuing contact with mental health services; (c) use of in-patient services; and (d) compulsory admissions. We included published papers that compared access to, or use of, primary or secondary mental health care in the UK by at least two ethnic groups with psychotic and non-psychotic mental disorders in primary and secondary care (see Appendix). We excluded specialist services for children, adolescents and older people, as well as forensic services and services for psychosexual disorders or substance misuse.

The following bibliographic databases were searched: ASSIA, CINAHL, the Cochrane Trials Register, EMBASE, HealthStar, Medline, PsycLIT, Science Citation Index and SIGLE. The search was restricted to studies published in English between January 1983 and October 2000 inclusive. We identified titles and abstracts of papers that potentially fulfilled our inclusion criteria. The grouped search terms included: (a) GREAT BRITAIN, HEALTH SERVICES, INDIGENOUS ETHNIC GROUP; (b) HOSPITALS, PSYCHIATRIC, MENTAL HEALTH SERVICES, PSYCHIATRY, MENTAL DISORDERS, PSYCHOTROPIC DRUGS; (c) ETHNIC or MULTI-ETHNIC, UK or BRITAIN or LONDON, MENTAL or PSYCHIATRY or PSYCHOTROPIC. In databases where MESH terms were available they were exploded and combined. Searches were adapted for the different databases and performed independently by two reviewers. These searches were supplemented by personal bibliographies of the investigators, forward citation tracking using the Science Citation Index and Social Science Citation Index databases and by references in retrieved articles. We sent a list of all included studies to three external experts, asking them to identify any other relevant studies.

Abstracts, if available, or full papers were assessed independently by two reviewers. All potentially relevant papers were then assessed against inclusion and exclusion criteria. Disagreement was resolved by a third reviewer. Data were extracted independently by two reviewers. Discrepancies were resolved by discussion or by a third reviewer. For studies reporting pathways to specialist mental health care, continuing contact with mental health services and use of in-patient services, we conducted a narrative review because differences in the measurement and reporting of outcomes made combination of point estimates inappropriate. The narrative analysis involved comparison of tabulated data and appraisal of methodological quality (see below). For the review of variation in compulsory admissions to in-patient facilities, in addition to a narrative review, we performed a meta-analysis to quantify variation between ethnic groups in compulsory admission to in-patient facilities. We have expressed the main results as combined odds ratios with the random effects method after performing tests for heterogeneity. We combined data on the proportions of White and Black patients who were compulsorily admitted in the primary studies. We examined publication bias and related biases in a funnel plot and carried out a test of funnel plot asymmetry. All analyses were performed with StatsDirect (http://www.statsdirect.com/).

Two reviewers independently assessed the methodological quality of each published study, with criteria adapted from Raine's review of gender differences in health service use (Table 1; Raine, 2000). We assessed the quality of published studies in four domains: sample source and size; method of ethnic group classification; adjustment for disease and confounding variables, including socio-economic status; and whether ethnic groups were combined appropriately in the analysis. Each of these was equally important in the allocation of quality score. Papers varied in their classification of ethnic groups and these were often aggregated into broad categories. In our analysis we have used the most commonly reported aggregated categories, comparing ‘Black’ and ‘South Asian’ groups with ‘White’ participants. We summarised the results of the papers and examined whether quality scores, geographical location of the research, the composition of the ethnic groups and study design could explain any heterogeneity of findings.

Table 1 Scoring system for methodological quality of paper

Sample source and size	Score	Adjustment for confounding variables	Score	Ethnicity categorisation
				Quality	Score	Use in the analysis	Score
Routine data (e.g. health authority, GP list data)	0	None	0	Third-party reports (e.g. ward staff categorisation, name-based methods, skin colour methods)	0	Inappropriate ethnic groups combined for major study outcomes (e.g. Blacks v. all others) or poor method of collecting ethnicity data	0
Project-specific data: <30 cases in ethnic groups for major outcomes	1	Age and/or gender	1	Self-reported ethnicity or use of census categories	1	Lumping of groups: reasonable combinations of groups collected by census/self-report method	1
Project-specific data: >30 cases in ethnic groups for major outcomes	2	Diagnosis or disease severity (give 1 point if this sample is selected by diagnosis)	1			All analysis done on ethnic groups without amalgamation, and self-report/census categories for categorisation	2
Project-specific data: total sample size >500	3	Comorbidity and risk factors for outcome of interest1	1-3
Maximum possible2	3		5		1		2

RESULTS

Primary studies identified and quality scores

A total of 545 potentially relevant titles and abstracts were identified. Sixty-seven papers were scrutinised. One duplicate publication was excluded, leaving 38 papers that met our criteria (see Appendix). One additional paper that was identified by an external reviewer did not meet our inclusion criteria. Details of each paper, including classification of ethnic groups, demographic characteristics, study design and quality ratings are summarised in Table 2 (the full review tables can be seen on the Institute of Community Health Sciences Queen Mary website at http://www.ichs.qmul.ac.uk/transcultural/TMS-Pub1.htm). Quality scores ranged from 0 to 11 and were summarised into three categories: low quality (0-3); medium quality (4-7); and high quality (8-11). There were five high-quality and 23 medium-quality studies. The total quality score was most highly correlated with adjustment for confounders (r=0.81) and methods for ethnic group classification (r=0.6). The more recently published papers achieved higher quality scores (r=0.57, P=0.0002).

Table 2 Characteristics of studies included in review (ethnic classifications and diagnoses are those used in the original studies)

Author (ethnic definition1)	Ethnic groups reported, n (%)	Actual sample size, n (% response rate)	Diagnoses, n (%)2	Sample source	Data source	Design and follow-up	Other variables in analysis	Quality
Bebbington et al, 1994 (D)	Black Caribbean 71	376	Psychosis 195	IP	Probably MR	CS, 1-day census and admission over subsequent 91 days	Age, gender, informal or formal admission, challenging behaviour, emergency or not sections	6
	White 305		Affective disorder 140
			Other 41
Bhugra et al, 1999 (D)	Asian 24	62 (100)	Schizophrenia	IP, OP, GP	MR, Int			5
	White 38
Bindman et al, 2000 (B)	White 49	97 (74)	SMI (all together)	Community sample of prevalent cases	MR, Int	CS, collected prospectively	Living alone, admission to hospital	6
	Black Caribbean 34
	Black African 11
	Asian 3
Birchwood et al, 1992 (D)	Indian 30	101 (73)	Schizophrenia 62	CMHT, IP	MR; contact with CMHT	CP, 12-month historical cohort	Age, gender, SES	3
	African—Caribbean 50		Psychosis NOS 30
	White English 74
	Irish 5
	Other 10
Burnett et al, 1999 (C)	African—Caribbean 38	100	Schizophrenia 100	IP, OP, GP, Pris, CMHT	MR, Quest, Int	2-year prospective data collection but a CS pathways study	Age, gender, SES	7
	White 38
	Asian 24
	African—Caribbean 183		Schizophrenia 809			20-year historical cohort using Camberwell Register
	White 626
Callan, 1996 (A)	African—Caribbean, M: 24, F: 26	134 (100)	Schizophrenia	IP	MR	Prospective CP, matched case—control	Age, gender, SES	7
	White British, M: 24, F: 17
Castle et al, 1994 (A+D)	White 79	484 (90)	Schizophrenia	IP, OP, CMHT	MR, Case register	FCPsych 19-year historical cohort	Age, gender, SES	4
	African—Caribbean 78		Paraphrenia
	Other		ONPsych
	UK, Eire, West Indies, Africa, Asia, Other
Cole et al, 1995 (B)	White 39	93 (88)	Schizophrenia 58	IP	Quest, Int	Retrospective CS	Age, gender, unemployment	8
	Black 38		Other 35
	Asian 16 & other
Commander et al, 1997b (B)	Asian 1516	GenPop 433 (28.9)	Schizophrenia 38.5/48.73	IP, OP, CMHT, GP.	MR, Quest, Int	CS, 6-months prospective admissions	Age, gender, SES	5
	Black 3068	GP 834 (74)	BAD 7.7/12.73	GenPop, For, Pris
	White 1195	OP psychiatric services 2451	Depression 12.3/11.83
			Neurosis 18.4/6.93
		IP specialist (including	Drug misuse 4.5/6.33
		For) 834	None 11.8/4.93
			Other 6.8/8.73
Commander et al, 1997c (B)	As above	As above	As above	As above	As above	As above	Age, gender	5
Commander et al, 1999 (B)	Asian 40	120 (100)	NAPsych (all)	IP	Int	Retrospective, CP at admission (120) and discharge (120)		5
	Black 40
	White 40
Crowley & Simmons, 1992 (D)	African—Caribbean/	152 (100%)	Psychosis 86	IP	MR	CS, Age- and gender-matched case—control	Age, gender, SES	2
	Asian 77		Organic 7
	White 75		Mania 10
			Neurosis 39
			Personality disorder 7
Davies et al, 1996 (A+B)	Black African 29	213 (93)	Schizophrenia 238	IP, OP, GP, CMHT	MR	CS, all prevalent cases of psychosis in a sector, retrospective	Age, number of admissions living alone	9
	Black Caribbean 115		APsych 13
	Black other 15		Other psychosis 188
	White 254		Total psychosis 439
Dunn & Fahy, 1990 (D)	White 165	268 (100)	Schizophrenia 88	IP	MR, Quest	CP	Age, gender	3
	Black 88		Hypomania 52
	Asian 7		Depression 11
	Uncertain 8		Drug-induced psychosis 24
			Personality disorder 22
Falkowski et al, 1990 (D)	African—Caribbean	100	Schizophrenia 45	IP who asbscond	MR, CS	CS, retrospective		2
	26		BAD 17
	African 8		PPsych 9
	UK/Irish 49		Personality disorder 8
	Other European 10		Neurosis 21
	Asian 6
	Other 1
Flannigan et al, 1994 (D)	White (66.4/76.8)4	CS 147	Psychosis	IP	MR	CP	Age, gender, SES	3
	Black Caribbean (22.9/15.8)4	Prospective 266
	Other (10.8/7.4)4
Glover et al, 1989 (A)	Jamaica	2466 (73.9)	All psychiatric, schizophrenia, mania, depression	IP	RD	Historical cohort using Mental Health Enquiry data over 3 years (1982-1985)	Gender	4
	Barbados
	Trinidad/Tobago
Goater et al, 1999 (B+C)	White 39	93 (73 at year 1, 71 at year 5)	Schizophrenia 35	IP, OP, CMHT, GP, AE, For, Pris	MR	Prospective incidence cases	Age, gender, SES	7
	Black 38		NAPsych 22		Quest
	Asian 11		APsych 22		Int
	Other 5
Gupta, 1991 (A)	Asian 86	172	Psychosis	IP, CP	MR	FCPsych, 5-year cohort	Age, gender	6
	White 86
Harrison et al, 1989 (D)	African-Caribbean 42	131 (100)	Psychosis 42	IP, OP, CMHT	MR, Quest, Int	CS, FCPsych	Age, gender	4
			RSch 28
	General population 89
Ineichen et al, 1984 (D)	White—UK 193	264 (100)	All diagnoses where there was compulsory admission	IP	MR	Retrospective, CS	Gender	4
	West Indian 43
	Other non-White 19
	Other White 9
Koffman et al, 1997 (B)	White (75)	3710 (93.2)	Schizophrenia	IP	MR, survey	CS, I-day census	Age, gender	5
	Black (15)
	Asian (4)
Lloyd & Moodley, 1992 (D)	Black 37	128 (95.2)	Schizophrenia 59	IP	MR, Int	One-day census of IP	Age, gender, compulsory detention	5
	Non-Black 101		BAD 31
			Other psychosis 11
			Neurosis 9
			Other non-Psych 15
			Unknown 9
McCreadie et al, 1997 (D)	Black Caribbean 131	468	Schizophrenia 468	IP, OP, GP, CMHT	MR, Quest	CS	Age, gender	4
	White 337
McGovern & Cope, 1987 (A+B)	White 511	587 (100)	Schizophrenia/PPsych 71	IP	MR	CS	Age, gender	5
	African—Caribbean 72
	British-born African—Caribbean 41		Affective disorder 56
			Cannabis psychosis 17
			Other 102
McGovern & Cope 1991 (A+C)	African—Caribbean 33	62 (87)	Schizophrenia	IP	MR	CS		4
	White 29
McKenzie et al, 1995 (A+C)	African—Caribbean 53	113	Psychosis	IP	MR, Int	Cohort		10
	White 60					Mean follow-up at 49 months (African—Caribbean) at 44 months (White)
	Total 191
Moodley & Thornicroft, 1988 (A)	West Indian 26	100 (100)	Schizophrenia 75	IP	MR, Quest combined	CS, CP	Age, gender, SES	2
	Other 9		Affective 165, Other psychosis 55
	White 65
Moodley & Perkins, 1991 (D)	African—Caribbean (48)	52 (86.7)	Schizophrenia (40.4)	IP	MR, Int	CP, CS	Age, gender, SES	3
	White (42)		BAD (25.5)
			Depression (27.7)
Odell et al, 1997 (B)	White	752 (72)	GHQ score	Practice lists	Int, MR	CS	GHQ score, gender of GP, psychiatric history, physical health	10
	African—Caribbean	125 GPs in 66 practices
	Asian
Owens et al, 1991 (C)	African—Caribbean 120	275 (100)	Schizophrenia and PSchizo 114	IP	MR, Int	Prospective	Age, gender, marital, PA, PCA	4
	White UK-born 120
			Depression 63
			Mania 41
			Other psychosis 23
			Neurosis, personality disorder, none 30
Parkman et al, 1997 (A+B)	White (UK-born) 114	184 (63)	Schizophrenia 59	CP in contact with SS, GPs, sheltered accommodation, voluntary, private, and self-help care; clergy; homeless and prison services	MR, Quest, Int	CS, of random sample of prevalent cases of psychoses in contact with specified agencies	Age, gender, SES	9
	White (non-UK-born) 20		APsych 7
			Other NAPsych 30
	Black Caribbean (UK-born) 27
	Black Caribbean (non-UK-born) 23
Perkins & Moodley, 1993 (D)	White (48)	52 (86)	Schizophrenia (40.4)	IP, CP	MR, Int	CS	Age, gender, PA, MHA, GPI, PI	3
	African—Caribbean (42)		BAD (25.5)
Singh et al, 1998 (B+C)	Black Caribbean 44	417 (86.7)	Psychosis 233 non-Psych 3	IP	Quest	CP, prospective	Age, gender, SES, psychosis, risk of violence	9
	Asian 21
	White 352
Takei et al, 1998 (A+D)	White 54	88 (81)	Schizophrenia, Non-schizo	IP admissions in 1973/4 followed up 18 years later	MR, Quest, Int	Prospective follow-up	Age, gender, SES, age at first onset	6
	African—Caribbean 34
Thomas et al, 1993 (A+C)	UK-born 1363	1692	All	IP	MR	CP, retrospective, CS	Age, gender, SES	5
	European		Schizophrenia
	African—Caribbean6 126		BAD
	African—Caribbean7 67
	Asian (Indian/Pakistani)6 62
	Asian7 74
Turner et al, 1992 (C)	African—Caribbean 50	100 (63)	Schizophrenia 75	IP	MR	Retrospective, prospective	Age, gender, SES	4
	African 13		Mania 25
	Caribbean 36		Personality disorder
	Asian 1		22
	European other 50		Psychosis 14
	UK White 38		Organic 16
	European 6		Depression 2
	Cypriot 1		Other 7
	Irish 5		None 8
	Unknown 12		Unknown 13
van Os et al, 1996 (B)	White 44	110	NAPsych 110	IP	MR	CS, retrospective	Age, gender	5
	African—Caribbean 29		(including 79 cases of RDC-based schizophrenia)
	African 32
	Other 5

Methodological findings

Most papers reported observational studies in health service settings using a cross-sectional design, without making a distinction between incident and prevalent cases (see Table 2). A total of 34 studies included information from medical records, 15 from direct interview and 10 used questionnaires. Clearly the majority of studies are therefore subject to recall bias and information bias, and where samples were recruited from in-patient and out-patient services, selection bias is likely to be important. Only three studies were of a prospective cohort design in which biased assessment of risk factors is minimised. As far as could be judged, potential biases due to study design did not explain heterogeneity. Thirty-five papers included Black participants; eight did not distinguish between people of African—Caribbean and African origin. Eight papers included South Asians as a separate category but did not distinguish between people of Pakistani, Indian or Bangladeshi origin. Six papers identified White subgroups but none analysed them separately. Ten papers used country of birth alongside ethnicity information. Nineteen studies did not use ethnic group census categories; studies collecting data before 1991 used proxy measures for ethnic group, such as place of birth. Thirteen papers did not give any explanation for their classification of ethnic groups. Most papers included small numbers of participants from each ethnic group. Twenty-eight papers adjusted for age, 28 for gender and 16 for some measure of socio-economic status, with 24 taking measures to address potential confounding due to variation in diagnoses or illness severity. Although socio-economic status is known to be a confounder in studies of ethnic variations in health, papers adjusting for it did not always score highly on the total quality scores. None of the papers reported sample size calculations. The majority of studies (26) investigated services in London; other locations were Birmingham (McGovern & Cope, 1987, 1991; Birchwood et al, 1992; Commander et al, 1997a , b , 1999), Manchester (Thomas et al, 1993), Nottingham (Harrison et al, 1989; Owens et al, 1991; Singh et al, 1998) and Bristol (Ineichen et al, 1984).

Do pathways to specialist care vary with ethnic group?

Black patients had more-complex pathways to specialist services, seeing at least three carers before contact with specialist services (Commander et al, 1999). Compared with White patients, a greater proportion of Black patients had some contact with a helping agency the week before psychiatric service contact (Harrison et al, 1989) and admission was more likely to follow a domiciliary visit (Commander et al, 1997a ). Compared with White and South Asian patients who visited their general practitioner (GP), Black people were less likely to be referred to specialist services (Thomas et al, 1993; Cole et al, 1995; Burnett et al, 1999). An explanation might be the GPs' lower likelihood of recognising a psychiatric problem in Black people (Odell et al, 1997; Bhugra et al, 1999; Burnett et al, 1999). However, among patients presenting to general practice who are recognised to have a mental health problem, Black patients were more likely to be found in specialist services (Commander et al, 1997b , c ). The police were more likely to be involved in admissions or readmissions of Black people (Thomas et al, 1993; Burnett et al, 1999; Commander et al, 1999). These three papers had medium-quality scores, as did one showing that police involvement before admission was explained by a lack of GP involvement rather than ethnic origin of the patients (Cole et al, 1995). Two medium-quality papers found that Black people were most likely to present in crisis, often seeing the duty psychiatrist in an accident and emergency department as a first point of contact with services (Turner et al, 1992; Cole et al, 1995). In contrast, a low-quality paper showed no variation between ethnic groups in the pathways to specialist care (Moodley & Perkins, 1991). Variation in the diagnostic categories and differences in ethnic group composition do not explain the heterogeneous outcomes of these studies.

In west London, specialist referral following primary care assessments appeared to be equally common among White and South Asian patients, but hospital admission was more likely among South Asians following a domiciliary visit (Bhugra et al, 1999; Burnett et al, 1999). In Birmingham, South Asians had the highest community rates of mental disorder, were the most frequent consulters in primary care and were less likely than White people to have their mental disorder recognised (Commander et al, 1997b , c ). Of all ethnic groups with a mental disorder, South Asians were the least likely to be referred to specialist care (Commander et al, 1997b ; Odell et al, 1997). Quality ratings were similar for papers from Birmingham and London and do not explain the different findings. However, the two papers reporting no variation in detection between South Asians and Whites contained smaller samples (total South Asians=24) than studies detecting significant differences (total South Asians=1516).

Is contact between specialist services and Black and South Asian people maintained?

Compared with White patients, services were less likely to maintain contact with non-Whites in one part of south London (Norwood) but not in another neighbouring area (Nunhead; McCreadie et al, 1997). Another paper from London found no variation between ethnic groups in continuity of community care (Bindman et al, 2000), whereas African—Caribbean people in Birmingham had more ‘broken contact’ with aftercare services (McGovern & Cope, 1991). All three of these papers were of medium quality with a cross-sectional design and specifically recruited African—Caribbean people. The study with the largest sample size found differences between geographical areas, suggesting that variation in local service configuration and practice are influential and that ethnicity alone does not account for variations in patient contact (McCreadie et al, 1997).

In two papers from London, compared with White people, Black people were more likely to be in contact with services at 5- and 18-year follow-up, respectively (Takei et al, 1998; Goater et al, 1999). These papers were of medium quality and both recruited Black people of Caribbean origin. There were no papers that looked specifically at the pattern of South Asian peoples' contact with specialist services.

Use of in-patient services

Of 20 papers on in-patients, 17 reported measures of in-patient service use (representation on in-patient units (13) consistently showed greater use of in-patient services by Black people (McGovern & Cope, 1987; Owens et al, 1991; Birchwood et al, 1992; Thomas et al, 1993; Flannigan et al, 1994; Callan, 1996; van Os et al, 1996; Commander et al, 1997b , c ; Koffman et al, 1997; McCreadie et al, 1997; Parkman et al, 1997; Takei et al, 1998). One of these positive studies was of high quality and nine were of medium quality. Two papers (of low quality) found no variation in in-patient service use (Ineichen et al, 1984; Castle et al, 1994). One paper, a first-incidence study, found an excess of in-patient use by White patients (Goater et al, 1999). A study of admissions found an excess of admissions among Black women in one London borough but a 40% lower admission rate compared with White women in another borough (Bebbington et al, 1994). Quality scores and differences in the ethnic composition of the samples, as far as this could be judged, do not account for this heterogeneity. The evidence most consistently suggests an excess use of in-patient facilities by Black patients. Two papers on first-contact (incident) admissions indicate no excess among Black people (Castle et al, 1994; Goater et al, 1999), despite adjustment for age, gender and socio-economic status in one paper (Goater et al, 1999). Among patients of Caribbean origin, Jamaicans had the highest annual admission rates to British hospitals (Glover, 1989).

Despite variation in study quality, geographical region and South Asian group composition and different degrees of adjustment for confounders and diagnosis, results for in-patient use were more consistent for South Asians than for Black patients. Compared with Black patients, South Asians were less likely to be admitted for in-patient care, had the lowest admission rates to secure wards (Commander et al, 1997c ; Koffman et al, 1997), had shorter admissions than other ethnic groups (Gupta, 1991) and were least likely to be readmitted (Birchwood et al, 1992). Compared with White patients, South Asians were more likely to be admitted to in-patient care (Commander et al, 1997b ; Koffman et al, 1997).

Compulsory admissions

Of 23 papers measuring compulsory admissions, the majority (18) showed a higher rate for Black compared with White patients (Ineichen et al, 1984; Moodley & Thornicroft, 1988; Harrison et al, 1989; Dunn & Fahy, 1990; Moodley & Perkins, 1991; Owens et al, 1991; Birchwood et al, 1992; Crowley & Simmons, 1992; Lloyd & Moodley, 1992; Perkins & Moodley, 1993; Thomas et al, 1993; Davies et al, 1996; Koffman et al, 1997; McCreadie et al, 1997; Parkman et al, 1997; Singh et al, 1998; Takei et al, 1998; Commander et al, 1999). Eight of these adjusted for age, gender and socio-economic status. Two adjusted for other potential confounders: class; past admissions; Mental Health Act status (compulsory admission for assessment or detention); police involvement; and general population representation (Perkins & Moodley, 1993; McKenzie et al, 1995). Other papers adjusted for age at first contact (Takei et al, 1998), psychosis, risk of violence and diagnosis (Singh et al, 1998). One paper that adjusted for first-contact admissions, past admissions and marital status showed excess admissions among Black people (Owens et al, 1991), challenging reports that the absence of a relative or intimate relationship accounts for compulsory admission (Cole et al, 1995). Davies et al (1996) recruited people with psychosis from a broad range of services. After adjustment for living alone as well as for age and total number of previous admissions, this paper still found excess compulsory admissions among Black people. Although papers reporting on in-patients suggest that a higher rate of readmissions and absconding explained the excess of compulsory admissions among Black people (Falkowski et al, 1990; Thomas et al, 1993), Davies et al (1996) adjusted for previous admissions and still found higher rates of compulsory admissions among Black people.

Some papers reported contradictory findings. For example, a medium-quality paper showed that, compared with White women, Black women were less likely to be voluntarily admitted in one London borough (South Southwark), but were more likely to be voluntarily admitted in another London borough (Hammersmith & Fulham; Bebbington et al, 1994). Such differences probably reflect variations in local practice and services. The majority of papers, using a variety of ethnic group measures and study designs and adjusting for a number of potential confounders, provide strong evidence for a relative excess of compulsory admissions of Black people.

Of three papers investigating compulsory admissions among South Asian patients, two low-quality papers found that these were less common among South Asian patients (Birchwood et al, 1992; Crowley & Simmons, 1992). One medium-quality paper found no difference between South Asian and White patients (Thomas et al, 1993). These findings are inconclusive as all three studies relied on routinely collected data from medical records and included small numbers of South Asian people.

We calculated a summary odds ratio from the 12 papers (Fig. 1) that reported the exact proportions of compulsorily admitted Black and White in-patients and sample sizes. The meta-analysis gave a pooled odds ratio (Blacks compared with Whites) of 4.31 (95% CI 3.33-5.58). Analysis of the funnel plot did not show evidence of publication bias (P=0.69).

Fig. 1 Odds ratio meta-analysis plot (random effects), Black v. White, from meta-analysis of 12 papers on compulsory admission to in-patient facilities.

DISCUSSION

Main findings

Of many studies addressing any one of our review questions, few actually used the same measure of outcome; therefore, we did not report estimates because, if presented in isolation, these few estimates are likely to give a biased impression. However, the original data tables are available on request and will be posted on the Queen Mary website. Our review suggests that Black people are overrepresented among in-patients, and that Asian patients use in-patient facilities less often than do White patients. Unexpectedly, two papers report an excess of voluntary admissions for White people. There is some evidence for variations in pathways to specialist mental health care, with Black people traversing more complex pathways. Variation in primary care assessments or primary care involvement could explain some of the ethnic differences in pathways to specialist services but the primary care literature is small. Despite concerns about Black people falling out of care, two papers showed that they were more likely to be in contact with services after 5 and 18 years. This could reflect effective delivery of necessary care, or professionals' anxieties about perceived risk. Black people on in-patient units were four times more likely to experience a compulsory admission compared with White people. This is consistent with research in forensic and prison services (Bhui et al, 1998; Coid et al, 2000) and with research informing the proposed amendments of the Mental Health Act. These reported twice the risk (Wall et al, 1999) and six times the risk of compulsory detention of Black people (Audini & Lelliott, 2002).

Explaining the findings

No papers reported investigations of discrimination as a risk factor. Perceived discrimination could have a detrimental effect on mental health, placing Black people at higher risk of mental illness and perhaps of mental health service use (Karlsen & Nazroo, 2002). Perceived discrimination within health care services may further compound this, whereas excessive use of compulsory admission could account for less satisfaction with services and fear of contact with services. Although social isolation (Cole et al, 1995) can also mediate higher contact with emergency services among Black patients, local variations in clinical practice and service provision are well recognised as sources of inequalities. Most studies did not compare regional variations of clinical practice or service configuration. Most studies were based in London. Even though the majority of the UK's minority ethnic communities live in the largest cities, mental health policy should reflect the needs of regions with smaller proportions of minority ethnic groups. Indeed, recent evidence suggests that contextual effects, such as a lower ethnic density, can actually lead to higher rates of schizophrenia requiring greater service use (Boydell et al, 2002).

Methodological considerations

The composition of any one ethnic group varied across the papers. This raises uncertainty about the generalisability of the findings from any single paper to populations that fall within the same ethnic group but have different ethnic subgroup profiles (for example, South Asians with different places of birth, country or region of origin within the Indian subcontinent). Improvement in the quality of the more recent publications is encouraging but the majority did not justify their classification of ethnic groups (Senior & Bhopal, 1994) and did not take account of possible variations in service use between subgroups within the larger ethnic categories. Adjustment for socio-economic status did not always relate to study quality, but adjustment for confounders generally and ethnic group definitions did relate to quality.

FUTURE PRIORITIES

Studies including White subgroups and South Asians, services outside of London and community and primary care studies were underrepresented in the literature. Future studies should explore explanations for variations that go beyond assigning all differences to ethnic origins. More-tailored study designs are needed to assess aetio-logical factors, mechanisms to explain ethnic variations and interventions. Where ethnic differences are evident, future investigations should address potential explanatory factors, such as discrimination, practice variation, service configuration and social isolation. Greater attention to place of birth, religion, duration of residence in the UK, place and level of education and cultural identity could also refine the interpretation of data beyond assigning all differences to ethnic origin.

Clinical Implications and Limitations

CLINICAL IMPLICATIONS

1. ▪ Black patients are more likely to be voluntary or compulsory in-patients.

2. ▪ Pathways to specialist services are more complex for Black patients, with more crisis routes of entry into care.

3. ▪ Primary care assessments show ethnic variations that could explain some of the variations in specialist service use.

LIMITATIONS

1. ▪ We excluded qualitative studies that might shed light on putative mechanisms that explain our findings.

2. ▪ The variety of study designs, quality of studies, ethnic group definitions and outcome measures makes comparison, and specifically meta-analyses, inappropriate. This limited the review to a narrative account of differences.

3. ▪ There is little information on White subgroups, practice and service variations, South Asian groups, services outside of London and primary care and community services.

APPENDIX

Inclusion and exclusion criteria

Inclusion criteria

Variation studies

UK-based studies, measuring variation in service use or provision for mental health care. Report data for more than one ethnic group, retrospective, prospective or cross-sectional studies of use or provision.

Intervention studies

International studies, including quantitative studies of interventions aiming to improve access and uptake of mental health services by different ethnic groups. The types of interventions could include educational (for users and/or health care professionals), organisational, financial, regulatory. Study designs could include controlled trial, parallel group or before-and-after studies.

Grey/unpublished literature

UK-based data sought from health authorities and by asking three ‘experts’ to identify any major omissions from the identified publications.

Service use

Studies that investigate the rates of use by ethnic groups of facilities provided for mental health care, whether service provision is targeted or non-targeted.

Service provision

Studies were included that investigate provision of services for different ethnic populations and the uptake of these services by their targeted population.

Management of disorders

Drug treatment of any affective or psychotic condition in primary or secondary care.

Participants

Adults identified by ethnic group with an affective or psychotic disorder.

Date of publication

January 1983 to end October 2000.

Language

English, French or German.

Exclusion criteria

Studies of services for children, adolescents, psycho-sexual disorders, substance misuse, alcohol misuse and forensic mental health were excluded.