Ethnic variations in compulsory detention under the Mental Health Act: a systematic review and meta-analysis of international data

1Ethnic minorities are subject to a disproportionate risk of involuntary psychiatric detention.,, Research has suggested that patients from black, Asian and minority ethnic (BAME) groups have more compulsory admissions to hospital,, longer inpatient stays, and more readmissions. However, other studies have reported no or weak associations between BAME populations and increased detention., Furthermore, the risk of involuntary psychiatric detention is not consistently higher in all or specific ethnic groups,,, potentially undermining actions to reduce inequalities and inform policy. Although evidence from the UK suggests increased detention under the Mental Health Act (MHA) for black populations,, less is known about south Asian (including Bangladeshi, Indian, Pakistani, Sri Lankan, and Nepali people),,, and migrant populations. Several explanations have been suggested for the observed disparities between minority and majority ethnic groups, such as increased prevalence of schizophrenia in some minority ethnic, and migrant, populations, insufficient patient awareness of mental health issues,, more frequent adverse experiences with mental health services, experience of racism or health-care provider discrimination,, and differing use of psychiatric services., However, few of these explanations provide adequate supporting evidence. A meta-analysis highlighted the importance of ethnic specificity in study design, considering black Caribbean patients separately and in addition to black patients more generally. However, there has otherwise been little evidence to systematically update and pool knowledge of the over-representation of BAME populations in those detained under the MHA in England since its revision and, to our knowledge, no systematic review or meta-analysis has been done from an international perspective. This study aims to expand upon existing systematic reviews and meta-analyses,, on the rates of detention for specific BAME populations from an international perspective, and outline explanations for any disparities between populations. Novel considerations of migrant populations are also reported.

2Evidence before this studyA large body of evidence suggests that black, Asian, and minority ethnic (BAME) groups are at an increased risk of compulsory detention under the Mental Health Act in England and Wales. However, there is marked variation in the composition of ethnic groups, definitions of ethnic groups, sample sizes, and reported magnitude of the effect sizes. The literature has typically focused on a small number of countries (primarily the UK) and has often neglected migrant populations. We searched MEDLINE (between Jan 1, 1946, and Nov 27, 2017), and PsycINFO (between Jan 1, 1806, and Nov 13, 2017), for systematic reviews and meta-analyses published in English with the search terms “minority groups” or “ethnic groups” or “BME” or “BAME” or “immigrants” or “refugees” and “Mental Health Act” or “commitment” or “admission” and “psychosis” or “schizophrenia” or “psychotic disorders”. We found no meta-analyses covering both the international and UK literature.Added value of this studyTo our knowledge, this is the first systematic review and meta-analysis to review both international and UK-based studies of compulsory detention, and the first to consider compulsory detention in migrant populations. This systematic review benefits from separate consideration of different ethnic groups where possible, avoiding grouping of culturally diverse populations. UK-based and international research showed significantly increased compulsory detention in several different ethnic minority and migrant populations, although UK research showed a more pronounced result.Implications of all the available evidenceOur findings support that compulsory detention and readmission in all BAME populations is significantly increased compared with majority groups, as is that of migrant populations compared with host nation populations. Detention rates vary across different BAME groups, with the highest rate seen in black Caribbean populations, and less marked, but still significantly increased rates in south Asian populations. Future research should aim to establish the causes of ethnic disparities in involuntary care and should avoid cultural stereotypes and assumptions. Amalgamation of ethnic groups should be discouraged to better inform policy and practice.

3Evidence before this study

4A large body of evidence suggests that black, Asian, and minority ethnic (BAME) groups are at an increased risk of compulsory detention under the Mental Health Act in England and Wales. However, there is marked variation in the composition of ethnic groups, definitions of ethnic groups, sample sizes, and reported magnitude of the effect sizes. The literature has typically focused on a small number of countries (primarily the UK) and has often neglected migrant populations. We searched MEDLINE (between Jan 1, 1946, and Nov 27, 2017), and PsycINFO (between Jan 1, 1806, and Nov 13, 2017), for systematic reviews and meta-analyses published in English with the search terms “minority groups” or “ethnic groups” or “BME” or “BAME” or “immigrants” or “refugees” and “Mental Health Act” or “commitment” or “admission” and “psychosis” or “schizophrenia” or “psychotic disorders”. We found no meta-analyses covering both the international and UK literature.

5Added value of this study

6To our knowledge, this is the first systematic review and meta-analysis to review both international and UK-based studies of compulsory detention, and the first to consider compulsory detention in migrant populations. This systematic review benefits from separate consideration of different ethnic groups where possible, avoiding grouping of culturally diverse populations. UK-based and international research showed significantly increased compulsory detention in several different ethnic minority and migrant populations, although UK research showed a more pronounced result.

7Implications of all the available evidence

8Our findings support that compulsory detention and readmission in all BAME populations is significantly increased compared with majority groups, as is that of migrant populations compared with host nation populations. Detention rates vary across different BAME groups, with the highest rate seen in black Caribbean populations, and less marked, but still significantly increased rates in south Asian populations. Future research should aim to establish the causes of ethnic disparities in involuntary care and should avoid cultural stereotypes and assumptions. Amalgamation of ethnic groups should be discouraged to better inform policy and practice.

9For this systematic review and meta-analysis, we included studies of samples in which two or more ethnic groups of any age were compared, and that compared the risk of compulsory inpatient psychiatric care in minority and majority ethnic groups. Included study outcomes were compulsory inpatient admission to hospital, compulsory inpatient readmission to hospital, and inpatient length of stay, and we considered only quantitative data.

10We developed our search strategy in consultation with an information scientist with experience in mental health, with a combination of keyword and subject heading searches. We searched MEDLINE (between Jan 1, 1946, and Nov 27, 2017), PsycINFO (between Jan 1, 1806, and Nov 13, 2017), Embase (between Jan 1, 1974, and Nov 20, 2017), Cochrane Controlled Register of Trials (between inception and Nov 27, 2017), and CINAHL (between Jan 1, 1981, and Nov 30, 2017).

11Although our search was not confined to countries in which black and other non-white ethnicities are minorities, all papers meeting inclusion criteria used either white or the dominant national group as their comparison group. Full search strategies are available in the appendix (pp 1–20).

12We contacted study authors; however, none contacted replied. We did not assess grey literature sources. Articles were translated to English, but none of the translated articles ended up being included. We sought summary estimate data rather than individual patient-level data.

13Two reviewers (EM and KA) independently screened all titles and abstracts identified and excluded studies that did not meet the inclusion criteria. Full articles were subsequently reviewed in duplicate and in cases of disagreement, consensus was achieved through the referral of a third senior reviewer (HM). An update search was done on Dec 4, 2018, to identify any additional papers published between Nov 1, 2017, and Dec 3, 2018. We supplemented the search strategy with a backward reference search of included studies and a forward citation search using Scopus. References for all included studies are available in the appendix (pp 44–46).

14The main change to our methods following protocol registration was that we expanded our inclusion criteria to encompass international literature in addition to literature from England and Wales. This change was made to provide an international context for our study, to further our understanding, and to include migrant populations, who are focused on more frequently in the international literature. Additionally, we did post-hoc meta-regression analyses to explore potential associations between predictor variables and ethnicity and to account for heterogeneity.

15Four independent reviewers (PB, EM, HG, and KA) extracted the data and all extraction was reviewed for accuracy. An electronic Microsoft Excel-based form was used to record data extraction. We planned to exclude studies which reported data already included in our dataset, but we did not find any duplicates.

16We calculated overall summary estimates (odds ratios [ORs]) and 95% CIs with a random-effects model using the R package metafor version 2.0. P<0·05 was considered to indicate a significant difference. We used ORs because most papers identified in our search either provided the number of events and sample sizes to calculate ORs, or gave statistics in the format of ORs when raw data were not provided.

17Studies varied in their specificity of classification of ethnic groups. Unlike previous studies, we attempted to avoid aggregate comparisons. Where possible, we organised studies into the subgroups black Caribbean, black African, south Asian, and east Asian. We classified studies solely reporting data for black or black, other groups as black, unspecified. We constructed a further non-specific classification of BAME, unspecified to contain studies reporting a mixture of minority ethnicities, for example, non-white British. We also did an analysis of studies comparing migrant groups (those born outside the host country) with host populations. We included only unadjusted data in the main analysis.

18We did four post-hoc meta-regressions with Comprehensive Meta-Analysis software (version 3) to explore possible causes of heterogeneity and to investigate differences between UK and international literature. With these meta-regressions, we examined possible predictors of the effect of ethnicity on compulsory admission and included mean age, proportion of women, publication year, and national context (England and Wales or internationally). We also did sensitivity analyses, including only studies rated as high quality or higher ethnic specificity for the primary outcome of compulsory admission.

19Four reviewers (PB, EM, HG, and KA) quality assessed the included studies. First, we applied the 14-item quality assessment checklist devised by Kmet and colleagues to each study. Each study was assessed against the 14 items using a 3-point scale with a score of 2 representing fully met, 1 representing partially met, and 0 meaning a study did not meet the criterion. A total score was calculated by adding up the scores achieved for each item. If a criterion was not applicable, it was excluded from the score calculation, and therefore from the maximum total score that could be achieved. A summary score (total sum divided by the total possible sum) was then calculated, representing the methodological quality of each study. These scores were calculated as a linear score from 0 to 100 and divided into three categories: low (≤49), moderate (50–74), or high (≥75) quality studies. Second, we assessed the quality of each study in terms of ethnic specificity with an adapted version of Raine's review of gender differences within health care, which was developed by Bhui and colleagues and has been used in previous similar reviews. Quality rating scores were between 0 and 14 and were categorised as follows: 0–3 (low), 4–7 (medium), and 8–14 (high). For both scales, quality assessment was discussed until a consensus was obtained and disagreements were resolved through consultation with two senior reviewers (SP and HM).

20We extracted data on study design, sample size, population type, country, diagnosis, age, sex, ethnicity, marital status, living status (living alone, with family, or with a significant other), education, occupation, the legal system or act (eg, whether it was the Mental Health Act, and if yes which version, or if a different country, which legal ruling the person was detained under, although this was rarely reported), and the previously mentioned study outcomes of interest (ie, involuntary admission to hospital, readmission to hospital, and inpatient bed days) and their associated statistical data. As in the study by Singh and colleagues, we extracted explanations for differences in psychiatric detention of BAME groups from included studies. We summarised these explanations and recorded any support from primary evidence (data from the paper itself). Unsupported explanations were those that were untested by the design of the study. We classified explanations into five domains adapted from Singh and colleagues: patient-related, illness-related, service-related, culture-related, and service-patient interface. If studies were previously summarised by Singh and colleagues, those explanations were retained. We reported explanations only when primary evidence for an association was identified.

21We assessed the degree of publication bias by visual examination of funnel plots.

22We calculated heterogeneity between studies with the I2 statistic. A value of 0% indicated no observed heterogeneity and 25%, 50%, or 75% tentatively signified low, moderate, or high heterogeneity between studies, respectively.

23This study follows the PRISMA guidelines and is prospectively registered with PROSPERO, number CRD42017078137.

24The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

25Our search identified 9511 studies for title and abstract screening, from which 296 potentially relevant full-text articles were identified (figure). Of these, 67 studies met inclusion criteria and were reviewed in depth. We also included an additional four studies after reference and citation searches. An updated search done on Dec 4, 2018, for studies published between Nov 1, 2017, and Dec 3, 2018, found no additional studies that met inclusion criteria. 1 953 135 participants were included in the studies we reviewed.

26Study characteristics are summarised in table 1. The 71 studies included in our review reported compulsory admission (63 studies), compulsory readmission (11 studies), and inpatient bed days (two studies). Two studies reported rate ratios, and one reported a risk ratio, which could not be pooled and included in our meta-analysis. Four studies reported adjusted data only so were also excluded from our main analysis.,,, Most studies reported routine data from specific hospitals or districts, although some compared population rates of admission. Studies were from high-income countries, and predominantly the UK (49 studies). Other countries represented were Canada (two studies), Italy (three studies), Ireland (two studies), the Netherlands (four studies), USA (five studies), Norway (one study), Switzerland (two studies), Denmark (one study), Spain (one study), and New Zealand (one study). We found high variability in study quality and scores awarded on the ethnicity checklist,, with scores ranging from 2 to 12. The main areas of bias centred on insufficient consideration of confounding variables. We examined funnel plots to investigate publication bias and found that studies were evenly distributed around the SE (appendix), suggesting that publication bias did not significantly affect our results. However, we observed high heterogeneity for all outcomes (table 2).

27Black ethnic groups were significantly more likely to be compulsorily admitted to hospital compared with white ethnic groups (black, unspecified OR 2·00, 95% CI 1·28–3·11, p=0·0022; black Caribbean 2·53, 2·03–3·16, p<0·0001; black African 2·27, 1·62–3·19, p<0·0001; table 2). Black Caribbean patients were also significantly more likely to be readmitted to hospital compared with white ethnic groups (2·30, 1·22–4·34, p=0·0102). We found no significant association between ethnicity and inpatient bed days (0·88, 0·18–4·19, p=0·8687), although this comparison included only two studies (table 2). People from Asian ethnic groups were significantly more likely to be compulsorily admitted to hospital compared with people from white ethnic groups (south Asian 1·33, 1·07–1·65, p=0·0091; east Asian 2·17, 1·47–3·22, p=0·0001). Only two studies reported compulsory inpatient readmission in south Asian patients and the results of these were not significant. Other minority ethnicities were significantly more likely to be compulsorily admitted to hospital compared with majority groups (1·66, 1·29–2·14, p<0·0001), as were migrant populations compared with host nation populations (1·50, 1·21–1·87, p=0·0003). Forest plots for our analyses are provided in the appendix (pp 25–30).

28Study location was a significant predictor of compulsory admission in black, unspecified groups, such that UK-based studies reported significantly increased odds of compulsory admission in black ethnic groups compared with international studies (table 3). The proportion of women in the sample was also a significant predictor of compulsory admission to hospital in black, unspecified, black Caribbean, and south Asian groups. This association remained significant when adjusted for age in black, unspecified and black Caribbean groups, but was no longer significant in south Asian groups. Publication date was a significant predictor of compulsory admission to hospital only in black Caribbean groups, and mean age was not a significant predictor of compulsory admission (table 3). Scatter plots of these data are provided in the appendix (p 31–35).

29We did secondary analyses on compulsory admission data to examine the effect of study quality on results (table 4). When including only studies scoring highly on the ethnicity checklist, results remained significant. When including only studies rated highly with the Kmet quality assessment scale, results remained significant in all black ethnic groups, but became non-significant in south Asian groups. Only six studies scored highly on both quality assessment scales, making these analyses difficult to draw conclusions from.

30We did a sensitivity analysis to investigate if studies excluded for reporting only adjusted data,,, could have significantly affected results (appendix p 36). We included three studies in the sensitivity analysis because they adjusted for demographic variables (age and sex) only.,, We observed only marginal differences, with no consequent difference in interpretation. One study adjusted for additional variables (car ownership and housing tenure) in addition to age and sex and reported only risk ratios; we excluded this study from the sensitivity analysis.

31We extracted explanations for disparities in psychiatric detentions from all papers included in the review (appendix pp 37–43). Of the 71 studies, 12 offered no explanation for differences in psychiatric detention of BAME groups, 21 solely offered explanations unsupported by primary evidence, and 38 offered at least one explanation supported by primary evidence (appendix pp 37–43). 24 classifications of explanations emerged over five domains (appendix p 37), of which ten were unsupported by any of the included literature, seven had a mix of supporting and contradictory primary evidence, and seven were supported by primary evidence. The most common explanations with supporting evidence included increased prevalence of psychosis, increased perceived risk of violence, increased police contact, absence of or mistrust of general practitioners, and ethnic disadvantages. By contrast, frequently mentioned unsupported explanations for disparities in detention included higher comorbid drug use in BAME groups, language barriers, poorer detection of mental illness, and greater stigma than in majority groups.

32This review expands on previous research on the use of involuntary psychiatric detention in ethnic minority communities, through examination of both UK and international data. Black ethnic groups (black Caribbean, black African, and black, unspecified) were more likely to be involuntarily admitted to hospital compared with those of white ethnicity. Black Caribbean individuals also had an increased risk of readmission to hospital. South Asian groups had a significantly increased risk of involuntary admission, as did east Asian patients, although interpretation was restricted by small study numbers. Our observed associations between compulsory admission to hospital and ethnicity remained significant when restricted to studies we assessed as being of high ethnic specificity. Following restriction of analysis to studies with a high methodological quality rating, only the south Asian association became non-significant. This result could suggest that less methodologically sound studies drove the reported higher risk for detention in south Asian populations. By contrast, continued significance when examining only studies with high ethnic specificity (regardless of other methodological aspects) contradicts this result. We could draw the most confidence from studies of both high methodological quality and clarity and consistency of ethnic classification, but only seven studies met these criteria.

33Both ethnically and culturally heterogeneous BAME, unspecified and migrant groups also had an increased risk of involuntary admission to hospital, which suggests that, although effect estimates were lower than some previous literature,, all minority populations in the countries studied, including migrants, are subject to increased risks of detention.

34Publication date predicted an association between black Caribbean ethnicity and involuntary care, with more recent studies reporting lower effect estimates. This result might reflect more rigorous study designs in recent literature. UK-based studies also showed a higher risk of compulsory admission to hospital for black, unspecified groups compared with international literature. Despite our attempts to provide specific ethnic classifications, the high heterogeneity for all groups could also reflect this variation in the UK and internationally. However, this effect is difficult to disentangle because the international studies were few and typically lacked the specificity of ethnic classification more common to British studies, preventing further post-hoc examination across all groups. The proportion of women in the sample strengthened associations in black Caribbean, black, unspecified, and south Asian groups. However, whole sample proportions cannot adequately describe intersectional experiences of race and sex and future robust investigation is required. Data on readmission to hospital and length of stay were scarce, and socioeconomic and clinical moderators by ethnic group and involuntary status were infrequently reported, preventing meaningful investigation.

35Of the 71 papers included in our systematic review, 34 (48%) offered no explanation for the variation in risk of detention among minority groups, or solely offered explanations without support from primary evidence. Untested explanations perpetuated in the literature largely dealt with lifestyle, cultural health beliefs, clinical characteristics, and demographic-bound assumptions of minority ethnic groups (eg, more drug use and greater community stigma of mental illness). Such untested hypotheses are of little use and are problematic when applied to aggregated and non-specific ethnic groupings, which often contain populations with varying lifestyles, health beliefs, culture, religion, and other demographic variables. Application of assumptions to combined groups, which fail to consider intersectionality in the perpetuation of risk,, precludes further inquiry into the range of risks to which these groups are subject. Likewise, explanations with supporting evidence, such as increased rates of psychoses in minority groups, require close examination. Psychosis alone is not a criterion for detention under the MHA, therefore the prevalence of psychoses in BAME communities is insufficient to explain ethnic inequalities in detention. The excess of detentions in BAME groups also applies to readmissions and not just those presenting with a first episode, so might be a function of the care experienced by those with established psychoses. Differences leading to mental illness trajectories that result in detention, such as perceived risk of violence, also warrant further investigation.,

36This study has several limitations. First, included studies examined psychiatric hospitalisation only. This investigation provides a good basis for examining inequalities in psychiatric care, but ethnic differences in other psychiatric contexts should also be examined through robust research. Additionally, the insufficient data on other aspects of care provided in most studies and the pooling of data in a meta-analysis cannot provide the necessary detail on the nature of differences in admissions, both for compulsory detention and treatment where the patient agrees to care. Coercive non-formal admissions can also happen, and perceived coercion can substantially affect a patient's experience of care. Furthermore, both civil and forensic commitment were combined in the analysis. This strategy allowed a broader inclusion of literature, but important differences between the two forms of compulsory hospital admission could have been missed. Similarly, important differences in legal systems in the different countries included in this systematic review should not be ignored. However, we believe that providing international data on ethnic disparity gives a clearer picture of the shortcomings of present research in tackling a global problem.

37To our knowledge, this systematic review and meta-analysis is the most comprehensive to date on ethnicity and involuntary psychiatric hospitalisation, integrating international comparisons and psychiatric detention of migrants. However, a substantial portion of the literature presented lacked the methodological quality to allow us to draw mechanistic or causal inferences from it. The included studies provide restricted information on socioeconomic, cultural, and structural determinants of detention, and integration of data on detention with such factors is an important area for further research. Retention of untested explanations in the studies covered by this systematic review might serve to entrench narratives of racial determinism and contribute little to a fuller understanding of the range of inequalities faced by minority ethnic groups who come into contact with psychiatric services. Research should prioritise longitudinal study designs that can investigate clinical, socioeconomic, and demographic contributions, and avoid simple techniques to analyse complicated problems—ethnicity is a complex construct that comprises multiple interacting variables. Research should also integrate qualitative assessment of service provider biases, group level stigma, or patient mistrust of health-care services to gain a thorough understanding of individual patient experience. Decision-making processes in psychiatric detention, which exclude patient and family input to risk management, should also be examined because their interaction with situational factors, such as available alternative treatment and under-resourced services, might reflect area deprivation experienced by BAME communities.

38Steve Gilbert: “Reading the research presented in this systematic review makes it clear that although research has been done with the best of intentions and hopes for change, it has not been adequately rooted in the people it is trying to help. Attempts to highlight injustices have backfired and given ammunition to those people who are not willing to stop and think—we have left our research too open to interpretation. We need to be asking different questions when considering study design—that is, we should seek not just to say what the state of play is, but also why the state of play is how it is. Another key question is, who is doing the research? Research can and should be done by a range of people who have followed different trajectories. Relevant to ethnicity and compulsory detention, experience of the criminal justice system, ethnic discrimination, or low-income households will positively contribute to the conclusions drawn. Furthermore, the methodology of research to date has felt distant—it does not reflect real experience. Data are good and useful to contribute to a fuller understanding, but have we become entrenched in the data? It is wrong to assume that if data say something is true then it must be true?”

39Steve Gilbert: “Reading the research presented in this systematic review makes it clear that although research has been done with the best of intentions and hopes for change, it has not been adequately rooted in the people it is trying to help. Attempts to highlight injustices have backfired and given ammunition to those people who are not willing to stop and think—we have left our research too open to interpretation. We need to be asking different questions when considering study design—that is, we should seek not just to say what the state of play is, but also why the state of play is how it is. Another key question is, who is doing the research? Research can and should be done by a range of people who have followed different trajectories. Relevant to ethnicity and compulsory detention, experience of the criminal justice system, ethnic discrimination, or low-income households will positively contribute to the conclusions drawn. Furthermore, the methodology of research to date has felt distant—it does not reflect real experience. Data are good and useful to contribute to a fuller understanding, but have we become entrenched in the data? It is wrong to assume that if data say something is true then it must be true?”

40This meta-analysis showed that all minority populations studied were subject to an increased risk of involuntary psychiatric detention. We are no closer to understanding or effectively addressing these ethnic inequalities in psychiatric care. Only research committed to well-designed longitudinal studies and multisectoral, intersectional approaches will be able to untangle the causes of this health-care inequality.