Study design and data collection

We conducted a prospective, multicentre cohort study, identifying all episodes of self-harm presenting to emergency departments in general hospitals in Manchester (three hospitals), Derby (two hospitals) and Oxford (one hospital). The study hospitals were chosen on a pragmatic basis – the centres included were those that had established monitoring systems. The cities of Manchester, Derby and Oxford have different profiles (Table 1), with ethnic groups forming a greater proportion of the general population in Manchester. According to the UK government's Index of Multiple Deprivation, where 353 local authority areas in England were scored on a number of indicators (covering a range of economic, social and health issues) into a single deprivation score, ¹³ Manchester was ranked fourth (worst), Derby sixty-ninth and Oxford one hundred and fifty-fifth.

Table 1

Sociodemographic profile of the three cities: 2001 Census data ¹⁸

Characteristic	England	Manchester	Derby	Oxford
% of population
Unemployed, 16–74 years	2	4	3	2
Lone parents, 16–29 years	5	8	6	3
No qualifications known, 16–74 years	26	28	28	18
‘Non-White’ – all ages	9	19	12	13
Black – all ages	2	5	2	3
South Asian – all ages	4	8	8	4
% of all South Asian people
Indian	51	18	49	40
Pakistani	35	71	50	45
Bangladesh	14	11	1	15

Data were collected using established monitoring systems in the three centres, described in full elsewhere. ^{Reference Hawton, Bergen, Casey, Simkin, Palmer and Cooper14,Reference Waters and Stalker15} Self-harm attendances were identified via detailed examination of computerised emergency department records and defined consistently across all three centres as intentional self-poisoning or self-injury, irrespective of motivation and degree of suicidal intent. ^{Reference Hawton and Catalan16} Most participants received a psychosocial assessment from emergency department staff and/or mental health specialists. For assessed participants, clinicians recorded a wide range of sociodemographic and clinical information using research assessment forms. For participants who were not assessed (for example, because they refused or took early discharge), basic information was collected by research clerks from medical records. In Manchester, data were collected for non-assessed individuals from computerised patient record systems from September 2002 onwards. Information on ethnicity was obtained where available and recorded either by the assessing clinician or from the hospital patient records system, using standard UK national 2001 census categories. We combined ethnic groups on a pragmatic basis. Our categories were in line with previous research, ^{Reference Bhugra, Desai and Baldwin17} namely: ‘South Asian’, including all people of Pakistani, Indian, or Bangladeshi origin; ‘Black’, including Black African–Caribbean or Black Other; ‘White’, including White British, Irish or White Other. We chose not to include those of ‘other ethnicity’ because of relatively low numbers and diverse composition of this miscellaneous category. An upper age limit of 64 years was also applied because of low numbers of older people who self-harmed from ethnic minority groups.

Data were analysed for a 6-year study period for Derby and Oxford (1 January 2001 to 31 December 2006). For Manchester, complete data for both assessed and non-assessed individuals were available from the 1 September 2002 and so data were analysed for just over 4 years (ending 31 December 2006).

Statistical analyses

The analyses were carried out using Stata version 10 for Windows. They were based on each individual's first presentation for self-harm during the study period. Two sets of analyses were performed.

First, self-harm rates per 1000 person-years were calculated for individuals aged between 16 and 64 years (this age range ensured consistency with the age bands in the 2001 Census data) ¹⁸ and with a postcode within the city catchment area of each of the hospitals in Manchester, Derby and Oxford. Approximate person-years at risk were generated by multiplying the ethnic group, gender and age-specific population estimates for each catchment area by the applicable study period for that centre. Incidence rate ratios (and their 95% confidence intervals) for ethnic groups compared with White groups, city, age and gender were calculated from Poisson regression models, with no significant evidence of overdispersion.

Second, differences between South Asian and Black people compared with White people were explored with respect to characteristics and clinical outcomes following the first episode of self-harm in the study period (index episode) for all individuals aged between 16 and 64 years who attended any of the study hospitals with self-harm, regardless of area of their residence. Statistical significance was assessed using chi-squared tests for method of self-harm, sociodemographic characteristics, precipitating factors and clinical characteristics. Analyses were performed separately for males and females. Analysis of clinical management outcomes (as percentages) following self-harm was conducted using log-binomial regression to estimate risk ratios, with variance estimates corrected for hospital clustering effects. Finally, 12-month rates of repetition of self-harm were calculated based on the proportion of individuals re-presenting within 12 months of index episode (excluding individuals without a full 12-month follow-up period). Log-binomial regression was used to examine differences in the risk of repetition.

Ethical approval

Oxford and Derby both have approval from local health/psychiatric research ethics committees to collect data on self-harm for local monitoring and multicentre projects. Self-harm monitoring in Manchester is part of a clinical audit system, and has been ratified by the local research ethics committee. All monitoring systems are fully compliant with the provisions of the Data Protection Act of 1998. All centres also have approval under Section 251 of the NHS Act 2006 (formerly section 60 of the Health and Social Care Act 2001) regarding the use of patient-identifiable information.

Rates of self-harm

Analysis of rates of self-harm was conducted for individuals of White, Black or South Asian ethnicity (n = 8401) resident within defined city population areas of each of the centres' catchment areas (Table 2). Rates of self-harm in young Black females were highest in all three cities. The pooled rate ratio for Black females aged 16–34 years compared with White females of the same age group was 1.70 (95% CI 1.46–1.98). When we conducted a conservative sensitivity analysis, ascribing all missing ethnicity data as relating to White persons, the rate ratio (RR) was attenuated but remained significantly elevated (RR = 1.21, 95% CI 1.04–1.40). Combining data across the cities, there was no difference in risk of self-harm in South Asian females aged 16–34 years compared with White females of the same age (RR = 0.99, 95% CI 0.87–1.11).

Table 2

Numbers, person-years denominators and self-harm rates per 1000 × city, gender, age and ethnic group^a

	White			South Asian			Black
	Self-harm (n = 7564)	Person years	Rate/1000	Self-harm (n = 499)	Person-years	Rate/1000	Self-harm (n = 338)	Person-years	Rate/1000
Manchester
Males
16–34 years	955	224 263	4.3	64	28 409	2.3	52	11 938	4.4
35–64 years	778	227 972	3.4	19	16 553	1.1	21	12 216	1.7
Females
16–34 years	1526	231 361	6.6	193	28 929	6.7	134	13 030	10.3
35–64 years	939	229 445	4.1	31	16 493	1.9	35	13 312	2.6
Derby
Males
16–34 years	504	148 362	3.4	35	18 828	1.9	12	3 102	3.9
35–64 years	374	215 448	1.7	9	14 514	0.6	9	5 148	1.7
Females
16–34 years	674	151 464	4.4	58	19 356	3.0	21	3 054	6.9
35–64 years	450	216 690	2.1	13	14 574	0.9	8	5 184	1.5
Oxford
Males
16–34 years	317	139 074	2.3	24	7 584	3.2	13	3 534	3.7
35–64 years	236	111 294	2.1	3	4 284	0.7	3	3 690	0.8
Females
16–34 years	544	136 530	4.0	41	7 776	5.3	25	3 948	6.3
35–64 years	267	110 496	2.4	9	4 536	2.0	5	3 624	1.4

We tested for differences in age/gender-specific rates between cities using a Wald chi-squared test for heterogeneity. Compared with White people, the rate ratios in South Asian and in Black people varied significantly by centre in young people aged 16–34 years (in males: χ² = 17.9, P = 0.001; females: χ² = 11.10, P = 0.03), that is, ethnicity did not have the same risk or protective effects in all cities. However, there was no evidence of heterogeneity in older people aged 35–64 years of either gender (males: χ² = 3.6, P = 0.46; females: χ² = 2.9, P = 0.57).

In comparison with White groups, there were significantly higher rate ratios for young Black females in all three cities (Table 3). In contrast, rate ratios in young Black males did not differ materially from those of young White males. There were lower rate ratios in older Black people of both genders compared with White people (Table 3), a difference that reached significance in Manchester. There was variation in relative risk in South Asian groups relative to White people, with significantly lower rate ratios in young people of both genders in Derby, a lower relative risk in males in Manchester, but no difference in either gender in Oxford. For older South Asian people relative rates were lower than in White people in both genders in all three cities, the differences being significant in Manchester and Derby.

Table 3

Age-specific rate ratios (RRs) for minority ethnic groups v. White people: by city

	South Asian	Black
	RRa (95% CI)	RRa (95% CI)
Manchester
Males
16–34 years	0.53 (0.41–0.68)	1.02 (0.77–1.35)
35–64 years	0.34 (0.21–0.53)	0.50 (0.33–0.78)
Females
16–34 years	1.01 (0.87–1.17)	1.56 (1.31–1.86)
35–64 years	0.46 (0.32–0.66)	0.64 (0.46–0.90)
Derby
Males
16–34 years	0.55 (0.39–0.77)	1.14 (0.64–2.02)
35–64 years	0.36 (0.18–0.69)	1.01 (0.52–1.95)
Females
16–34 years	0.67 (0.51–0.88)	1.55 (1.00–2.39)
35–64 years	0.43 (0.25–0.75)	0.74 (0.37–1.50)
Oxford
Males
16–34 years	1.39 (0.92–2.10)	1.61 (0.93–2.81)
35–64 years	0.33 (0.11–1.03)	0.38 (0.12–1.20)
Females
16–34 years	1.32 (0.96–1.82)	1.59 (1.06–2.37)
35–64 years	0.82 (0.42–1.60)	0.57 (0.24–1.38)

Characteristics

Analyses of sociodemographics, precipitating factors and clinical characteristics were conducted for individuals of White, Black or South Asian ethnicity×gender regardless of area of residence (online Tables DS1 and DS2). In both genders, ethnic minority groups (Black and South Asian) were younger than the White groups, and less likely to have clinical characteristics known to increase risk of further suicidal behaviour compared with their White peers (that is, alcohol use within 6 h of the self-harm episode, previous self-harm, history of psychiatric treatment).

Clinical management and outcome

Analyses of clinical management following self-harm showed that young Black females were less likely to receive a specialist psychiatric assessment compared with White females (Table 4). Ethnic minority groups of both genders were less likely to present to the emergency departments within the study hospitals with further self-harm (Tables 4 and 5). We fitted extra models adjusting for age in both genders but estimates were materially unaltered.

Table 4

Clinical management and repetition following self-harm: Black and South Asian v. White females aged 16–64 years^a

	White (n = 7938)		South Asian (n = 459)				Black (n = 288)
Clinical management outcome	n	%	n	%	RR	95% CI	n	%	RR	95% CI
Assessmentb
Specialist psychiatric assessment	5756	73	289	63	0.9	0.7–1.03	186	65	0.9	0.8–0.98
Clinical managementb,c
Psychiatric referral	2651	33	81	18	0.5	0.5–0.6	60	21	0.6	0.5–0.8
Alcohol or drug services referral/told to see	344	4	2	0.4	0.1	0.1–0.2	10	3	0.8	0.3–2.1
Other services referral/told to see	1893	24	97	21	0.9	0.6–1.3	59	20	0.9	0.6–1.2
General practitioner referral	4710	59	262	57	1.0	0.8–1.1	154	53	0.9	0.9–0.98
No formal follow-up	1555	20	142	31	1.6	0.98–2.5	97	34	1.7	1.00–2.9
Self-harm repetitiond	1646	24	49	13	0.5	0.4–0.7	31	13	0.6	0.4–0.8

Table 5

Clinical management and repetition following self-harm: Black and South Asian v. White males aged 16–64 years^a

	White (n = 5949)		South Asian (n = 218)				Black (n = 145)
Clinical management outcome	n	%	n	%	RR	95% CI	n	%	RR	95% CI
Assessmentb
Specialist psychiatric assessment	4292	72	147	67	0.9	0.8–1.1	99	68	0.9	0.9–1.01
Clinical managementb,c
Psychiatric referral	1862	31	58	27	0.9	0.7–1.1	42	29	0.9	0.7–1.3
Alcohol or drug services referral/told to see	530	9	13	6	0.7	0.3–1.5	8	6	0.6	0.2–1.7
Other services referral/told to see	1324	22	38	17	0.8	0.5–1.2	33	23	1.0	0.6–1.7
General practitioner referral	3301	55	119	55	1.0	0.8–1.2	68	47	0.8	0.8–0.9
No formal follow-up	1232	21	61	28	1.4	1.02–1.8	39	27	1.3	0.9–1.8
Self-harm repetitiond	1223	23	29	16	0.7	0.5–0.97	19	15	0.7	0.4–0.99

Main findings

Our main result was that across all three cities young Black females were at increased risk of self-harm. We also found differences in clinical management, with BME groups being less likely to receive a specialist psychiatric assessment and psychiatric follow-up services than the White population. In addition both minority ethnic groups in the older age range had a lower risk in all cities, and there was a variation in rates between cities among young South Asian males and females.

To our knowledge this is the first study to show significantly higher rates of self-harm in young Black females across a number of cities using large population based databases. A previous study in the UK over 20 years ago did suggest that there may be an elevated risk of self-harm in this group, ^{Reference Merrill and Owens19} although these findings were limited by small sample size and data confined to one hospital. A more contemporary study among people with recent previous contact with psychiatric services showed a higher standardised mortality ratio for suicide in Black African and Caribbean females aged 25–39 compared with their White peers. ^{Reference Bhui and McKenzie4} Our finding of lower rates of self-harm in older males and females in both ethnic minority groups are consistent with previous studies. ^{Reference Cooper, Husain, Webb, Waheed, Kapur and Guthrie6,Reference Bhui, McKenzie and Rasul11} We did not find the high rates of self-harm previously reported in young South Asian females ^{Reference Bhui, McKenzie and Rasul11} across the centres in our study.

Interpretation of findings

Several of our findings require further explanation. High rates in some BME groups may be explained by characteristics that confer greater vulnerability. Poor outcome may be influenced by complex socioeconomic factors ^{Reference Stronks and Kunst20} or may be culturally specific with differences between age and gender groups. We lacked the necessary explanatory variables to test these hypotheses. However, young Black females who self-harm may be experiencing greater social adversity, as in our cohort they were more likely to be unemployed and report housing problems compared with White women. We found that people from ethnic minority groups were more likely to be students than their White counterparts. Academic pressure may also have contributed to increased rates of self-harm, especially in women. ^{Reference Garlow, Rosenberg, Moore, Haas, Koestner and Hendin21,Reference Verger, Combes, Kovess-Masfety, Choquet, Guagliardo and Rouillon22} Previous studies examining high rates of psychosis in BME groups, particularly in young Black–Caribbeans, suggest that socioeconomic factors contribute part of the explanation. ^{Reference Karlsen and Nazroo23,Reference Kirkbride, Barker, Cowden, Stamps, Yang and Jones24}

The lower rates of self-harm in young Black males across centres deserves further investigation. Previous studies have found increased risk of mental illness and higher rates of completed suicide in this group compared with the White reference group. ^{Reference Bhui and McKenzie4,Reference Sharpley, Hutchinson, Murray and McKenzie25} Contemporary mainstream hip-hop epitomises ‘Black’ youth culture and places emphasis on strength, aggression and virility. However, stereotypical ‘manly’ behaviour may mean that young Black males do not seek help for emotional problems. A recent community study of suicidal behaviour in the UK provides some support for this interpretation of our results. ^{Reference Crawford, Nur, McKenzie and Tyrer26} These researchers found limited evidence of higher levels of suicidal ideation in second-generation immigrants and that ethnic minority groups were half as likely to seek medical attention following self-harm compared with White groups.

In our cohort older ethnic minority people of both genders had lower rates of self-harm than their White counterparts in all three cities (although only significantly so in Manchester and Derby). One explanation is that they have lower rates per se. Another is that older BME groups may be more reluctant to present to statutory services. In an American study, African Americans were less likely to access specialist mental health services compared with non-Hispanic White people. ^{Reference Garland, Lau, Yeh, McCabe, Hough and Lansverk27} Suggested reasons for a resistance to presentation to hospital following self-harm have included that seeking help for mental distress is considered stigmatising and socially unacceptable, and that services were not accessible or considered relevant. ^{Reference Crawford, Nur, McKenzie and Tyrer26,Reference Ahmed, Mohan and Bhugra28}

Clinical management is guided by knowledge of risk factors from epidemiological studies in the context of an individual patient presentation. An explanation of the less frequent specialist psychiatric assessment received by BME groups could relate to their ‘low risk’ clinical characteristics; for example, they were less likely to live alone, use alcohol with the self-harm attempt and have a previous psychiatric history or history of self-harm. Some BME groups were slightly less likely to be subsequently referred for formal follow-up. In addition to having apparent low-risk social and clinical characteristics, a further explanation for low-risk management might be related to how individuals from different ethnic groups communicate distress. In a recent UK study on response to childbirth, Black females had lower depression scores than White women, despite experiencing greater social adversity. ^{Reference Edge and Rogers29} The ‘discourse of strength’ (a perceived ability to deal effectively with a range of problems) attributed to young Black females may be part of their sense of identity, and admitting to depression therefore a sign of weakness. ^{Reference Edge30} They may still actually experience psychological distress and resort to self-harming behaviour at a time of crisis, but this attitude of being strong may mean that these young females do not subsequently communicate risk to clinical staff.

The rate ratio for repetition of self-harm was significantly lower in all ethnic minority groups compared with White groups. This may be a result of reduced risk as they were much less likely to have those characteristics known to increase risk of suicidal behaviour. ^{Reference Kapur, Cooper, King-Hele, Webb, Lawlor and Rodway31} However, it may also be explained in terms of disillusionment with statutory services. Black and minority ethnic groups are generally perceived to have poor experiences of mental health services. ³² It is possible that on initial presentation some ethnic minority groups may not have received appropriate help, which may have affected their willingness to re-present. ^{Reference Raleigh, Irons, Hawe, Scobie, Cook and Reeves33,Reference Taylor, Hawton, Fortune and Kapur34}

Previous studies have found an increased risk of self-harm in young South Asian females compared with White women. ^{Reference Bhui, McKenzie and Rasul11} We did not find this. There are a number of possible interpretations for the difference in results in young South Asian groups compared with previous findings and between centres. First, previous studies were carried out in Birmingham and London ^{Reference Bhugra, Desai and Baldwin17,Reference Merrill and Owens35} and not all South Asian populations are the same. For example within our three centres the proportions of the population that were of Indian, Pakistani and Bangladeshi origin varied considerably (Table 1). Our results might therefore reflect the particular mix of South Asian populations in the three centres. Second, comparative studies took place over 10–25 years ago and the young British South Asian population is likely to have changed over time. ^{Reference Platt, Simpson and Akinwale36} Third, since these findings, considerable attention has been drawn to this problem, with consequent improvement of services. ^37,38

The higher rate ratios observed in young South Asian people compared with young White people in Oxford in relation to the other cities might be a result of differences in the relative socioeconomic profiles between ethnic groups in different cities. Differences in the relative density of ethnic groups between cities may also explain this variation in rates. Oxford has the lowest proportion of South Asian people within its population compared with the other cities, although it had higher rates of self-harm in South Asian young males and females compared with their White counterparts. This could reflect the cultural incongruence and ecological-effect modification suggested in previous research – a negative correlation between the incidence of psychological distress and the size of an ethnic group relative to the total population. ^{Reference Neeleman and Wessley8,Reference Bhugra and Arya39} Unfortunately, we had no measure of these contextual variables to enable examination of their potential effects.

Strengths and limitations

The strength of this study is that we used large, contemporary population-based databases from three separate centres and examined the three main ethnic groups living in the UK. The data were collected principally from urban populations and this might limit the generalisability of the findings, although the populations covered by the three centres had different socioeconomic characteristics. ^{Reference Hawton, Bergen, Casey, Simkin, Palmer and Cooper14} The ethnic minority categories we applied were broad and did not take account of cultural identity. ^{Reference Bhui40} However, it enabled estimation of precise rates and relative risks stratified by age and gender and allowed direct comparison with previous research on ethnic groups in the UK. This may have concealed differences between ethnic groups within the categories we used. All three centres may have overassigned people to ethnic minority groups compared with Office for National Statistics ascertainment procedures. Ethnicity was recorded by healthcare staff who may or may not have asked the individual to categorise themselves. Ethnicity was recorded for 75% of individuals. Ethnicity is not recorded comprehensively in hospital settings, although our capture rate was significantly higher than a recent survey of users of community mental health services in England. ^{Reference Raleigh, Irons, Hawe, Scobie, Cook and Reeves33} Even so, there is the potential for selection bias. However, when we conducted a conservative sensitivity analysis on rates of self-harm on White versus Black females aged 16 to 34 years ascribing all missing data to the White groups the conclusion was essentially unchanged.

Implications for services and further research

Those designing services for people who self-harm need to be aware of the different levels of risk of self-harm and the variations in risk characteristics in different ethnic groups. Services also should be able to respond to the varied needs within these groups. It may be that a failure of professionals to recognise cultural factors at play, and an ignorance of available services, contribute to the lack of recognition of mental health problems and subsequent failure to offer (and for ethnic minority patients to engage in) further services. The challenge is to make services more culturally sensitive. There is some evidence for effectiveness of cultural competency training in demonstrating a change in skills and attitudes of clinicians. ^{Reference Bhui, Warfa, Edonya, McKenzie and Bhugra41} Future culturally sensitive studies might help us achieve a greater understanding of the suicidal process in ethnic minority groups.