Ecological studies

Among the ecological studies, all analyses examined the association between ethnic density and area-level rates of hospital admissions. Five studies looked at general psychiatric admissions, ^{Reference Klee, Spiro, Bahm and Gorwitz19–Reference Levy and Rowitz21,Reference Muhlin23,Reference Rabkin24} two looked at admissions for schizophrenia, ^{Reference Cochrane and Bal18,Reference Mintz and Schwartz22} and one also looked at admissions for bipolar disorder. ^{Reference Mintz and Schwartz22} There was great variability between the studies in population, study design and methods (online Table DS1). All but one study (published in 1988) ^{Reference Cochrane and Bal18} were published before 1980 and only one ^{Reference Mintz and Schwartz22} adjusted for area-level socioeconomic status. ^{Reference Mintz and Schwartz22}

FIG. 1 Flow diagram showing review process.

As can be inferred from Fig. 3, US studies provided the most consistent support for an ethnic density effect. ^{Reference Klee, Spiro, Bahm and Gorwitz19,Reference Levy and Rowitz21–Reference Rabkin24} All found a protective association of ethnic density with admission rates among the ethnic minority groups sampled in the studies, which included Black ^{Reference Levy and Rowitz21,Reference Rabkin24} and Puerto Rican people, ^{Reference Rabkin24} groups of European origin ^{Reference Mintz and Schwartz22,Reference Muhlin23} and the ‘non-White’ population in general. ^{Reference Klee, Spiro, Bahm and Gorwitz19} Outside the USA, findings are less consistent. A UK-based study of ethnic density, exploring hospital admissions for schizophrenia of four different ethnic groups (Irish, Indian, Pakistani and Caribbean) found a significant protective association with higher ethnic density only for Irish men. ^{Reference Cochrane and Bal18} Similarly, an Australian study of people of Greek and Italian origin in three time periods found a protective association with higher same-ethnic density only for Greek women between 1965 and 1970. ^{Reference Krupinski20} However, in the Australian study, the threshold for high same-ethnic density was >5%, and the UK study did not report ethnic density levels. It is, therefore, possible that neither study included samples living at same-ethnic densities high enough to offer any health benefit.

TABLE 1 Summary of measures used in included studies

	Studies, n
Ecological hospital admission studies
General psychiatric admissions	5
Admission for schizophrenia	2
Admission to hospital for bipolar disorder	1
Depression and anxiety
Depression (items from Center for Epidemiologic Studies Depression Scale)	9
Diagnostic Interview Schedule/DSM-III	1
General Health Questionnaire – 12	3
Clinical Interview Schedule – Revised, neurotic symptoms	2
Rumbault's Psychological Well-Being Scale	1
Self-report of clinical diagnosis	1
Self-report of sad or depression	2
Psychoses
Clinical assessment of multiple sources	4
Psychosis Screening Questionnaire	1
Suicide and self-harm
Suicide as most probable cause of unnatural death	1
Hospital attendance for self-harm	1

FIG. 2 Bar chart showing the number of independent analyses by direction of association and sample size of the analysis.

Depression and anxiety

Sixteen data-sets investigated the association between ethnic density and rates of depression or anxiety ^{Reference Abada, Hou and Ram25–Reference Yuan43} (online Table DS2). All these studies used multilevel data, however, four studies did not account for the multilevel structure or clustering of the data. ^{Reference Ecob and Williams28,Reference Halpern and Nazroo29,Reference Oliver33,Reference Shields and Wailoo37,Reference Ying and Akutsu42}

FIG. 3 Tree map showing all independent analyses from ecological studies as boxes, labelled with reference number of paper of origin, scaled by sample size and shaded by result of ethnic density association (details in text).

Among the 11 North American studies, 10 were of US populations, ^{Reference Aneshensel, Wight, Miller-Martinez, Botticello, Karlamangla and Seeman26,Reference Henderson, Diez Roux, Jacobs, Kiefe and Williams30–Reference Ostir, Eschbach, Markides and Goodwin34,Reference Tweed, Goldsmith, Jackson, Stiles, Rae and Kramer38–Reference Wight, Cummings, Karlamangla and Aneshensel41,Reference Yuan43} and 1 Canadian. ^{Reference Abada, Hou and Ram25} The majority of these studies investigated Black and Hispanic samples (Fig. 4). Of these North American studies that examined the association between overall-ethnic density and depression, ^{Reference Abada, Hou and Ram25,Reference Henderson, Diez Roux, Jacobs, Kiefe and Williams30,Reference Oliver33,Reference Wickrama, Noh and Bryant39,Reference Wight, Aneshensel, Botticello and Sepulveda40} two reported a significant association. Both of these were studies of adolescents aged between 12 and 19. ^{Reference Abada, Hou and Ram25,Reference Wickrama, Noh and Bryant39,Reference Wight, Aneshensel, Botticello and Sepulveda40} In one of these studies, based in Canada, overall-ethnic density had an adverse association with the mental health of ‘visible minority’ adolescents. ^{Reference Abada, Hou and Ram25} A visible minority person is defined under Canada's Employment Act as being a person who is neither White nor of aboriginal origin. ^{Reference Abada, Hou and Ram25} In the other study, based in the USA, overall-ethnic density also had non-significant adverse associations with the number of depressive symptoms of Hispanic and Asian/Pacific Islander adolescents, ^{Reference Wight, Aneshensel, Botticello and Sepulveda40} but among African American adolescents, this association was reversed – overall-ethnic density was protective against depressive symptoms. ^{Reference Wickrama, Noh and Bryant39,Reference Wight, Aneshensel, Botticello and Sepulveda40}

Among US adults, there is evidence that same-ethnic density had a protective association with depression and anxiety. Two ^{Reference Tweed, Goldsmith, Jackson, Stiles, Rae and Kramer38,Reference Yuan43} out of seven studies of same-ethnic density among Black people in the USA found such protective associations. However, one of these studies found the reverse association. ^{Reference Mair, Diez Roux, Osypuk, Rapp, Seeman and Watson32} Two ^{Reference Ostir, Eschbach, Markides and Goodwin34,Reference Ostir, Eschbach, Markides and Goodwin34} out of five studies of same-ethnic density among Hispanic people found protective associations. In addition, there is some evidence that ethnic density may protect against depression for Hispanic people living in socioeconomically disadvantaged communities. ^{Reference Wight, Cummings, Karlamangla and Aneshensel41} A possible explanation for the lack of consistent protective associations for Hispanic people is sample size; two of the studies of Hispanic people included very small samples at the individual level, ^{Reference Aneshensel, Wight, Miller-Martinez, Botticello, Karlamangla and Seeman26,Reference Wight, Cummings, Karlamangla and Aneshensel41,Reference Yuan43} and a third study contained a large sample of individuals, but few areas. ^{Reference Wight, Aneshensel, Botticello and Sepulveda40} It is less clear why there is some discrepancy in the findings among Black people, especially regarding the adverse association found in one study ^{Reference Mair, Diez Roux, Osypuk, Rapp, Seeman and Watson32} that contrasts with other studies that found protective associations.

FIG. 4 Tree map showing all independent analyses from studies of depression and anxiety as boxes, labelled with reference number of paper of origin, scaled by sample size and shaded by result of ethnic density association (details in text).

The majority of UK-based studies investigated Black or South Asian samples (Fig. 4). Two studies that combined ethnic minority groups found protective associations with overall-ethnic density. ^{Reference Das-Munshi, Becares, Dewey, Stansfeld and Prince27,Reference Halpern and Nazroo29} Another similar study found that the adverse association between overall-ethnic density and common psychiatric symptoms among White people was not found among a mixed sample of ethnic minorities, although it was not clear whether the association was reversed. ^{Reference Propper, Jones, Bolster, Burgess, Johnston and Sarker36}

Among studies that looked specifically at South Asian people, a study in Glasgow found that South Asian people were marginally less likely to have experienced depression and had fewer psychiatric symptoms when living in areas of higher same-ethnic density. ^{Reference Ecob and Williams28} However, studies that looked at specific groups within the South Asian category revealed some inconsistencies. For example, one study found a significant protective association with same-ethnic density for people of Indian origin, a non-significant protective association for Bangladeshi and African–Asian people, and a non-significant adverse association for people of Pakistani origin. ^{Reference Halpern and Nazroo29} In another study, protective associations were found for Bangladeshi, but not for Pakistani or Indian people. ^{Reference Das-Munshi, Becares, Dewey, Stansfeld and Prince27} In yet another study, protective associations were found for Pakistani and Indian mothers, but an adverse association was found among Bangladeshi mothers. ^{Reference Pickett, Shaw, Atkin, Kiernan and Wilkinson35} However, in this latter study, a very high proportion of the Bangladeshi sample was living in the most deprived areas in the UK. This may have rendered statistical adjustment for area deprivation less effective, or alternatively, the finding may indicate a deprivation threshold beyond which protective ethnic density effects are no longer important.

Four papers investigated outcomes related to depression and anxiety among Black people in the UK. Only one of these studies found that ethnic density was protective. ^{Reference Halpern and Nazroo29} It must be noted, however, that investigations of the ethnic density effect among Black people in the UK should not be directly compared with US studies. This is because the levels of Black ethnic density in the UK are typically very low, in contrast to US levels. If there is such a threshold of ethnic density where protective effects start to be measurable, it may be the case that UK levels for Black people are below this threshold.

Psychoses

Five studies investigated the association between ethnic density and psychoses, ^{Reference Halpern and Nazroo29,Reference BeAcares, Nazroo and Stafford44–Reference Veling, Susser, van Os, Mackenback, Selten and Hoek49} using four data-sets from the UK and one from The Netherlands (online Table DS3). All but one study ^{Reference Schoefield, Ashworth and Jones48} included analyses of the association between overall-ethnic minority density and psychoses among a combined sample of ethnic minority groups. Two studies ^{Reference Kirkbride, Morgan, Fearon, Dazzan, Murray and Jones46,Reference Kirkbride, Boydell, Ploubidis, Morgan, Dazzan and Mckenzie47,Reference Veling, Susser, van Os, Mackenback, Selten and Hoek49} also investigated overall-ethnic density associations for specific ethnic minority groups, and another two ^{Reference Halpern and Nazroo29,Reference BeAcares, Nazroo and Stafford44,Reference Schoefield, Ashworth and Jones48} investigated same-ethnic density associations for specific ethnic minority groups.

Overall-ethnic density was consistently protective for the combined ethnic minority samples, although in one study, the protective associations were restricted to schizophrenia and not found among non-affective psychoses. ^{Reference Kirkbride, Morgan, Fearon, Dazzan, Murray and Jones46,Reference Kirkbride, Boydell, Ploubidis, Morgan, Dazzan and Mckenzie47} The two studies that investigated overall-ethnic density associations for specific groups had different results. In the UK study no ethnic density associations were found for any of the specific ethnic minority groups, which included ‘Black – Caribbean’, ‘Black – African’, ‘Asian’, ‘Mixed ethnicity’, ‘White – other’ and ‘other ethnic groups’. ^{Reference Kirkbride, Morgan, Fearon, Dazzan, Murray and Jones46} However, in the Dutch study, higher overall-ethnic density was protective for psychoses among those of Moroccan ethnicity but not those of Surinamese or Turkish origin. ^{Reference Veling, Susser, van Os, Mackenback, Selten and Hoek49} The studies that investigated same-ethnic density associations for specific ethnic minority groups found similar results. ^{Reference Halpern and Nazroo29,Reference BeAcares, Nazroo and Stafford44,Reference Schoefield, Ashworth and Jones48} Ethnic density was found to be protective for people of Indian and Bangladeshi origin, ^{Reference Halpern and Nazroo29,Reference BeAcares, Nazroo and Stafford44} as well as for Black people. ^{Reference Schoefield, Ashworth and Jones48} The one contrasting finding was that, in one study, ethnic density was associated with increased risk of psychotic symptoms among Pakistani people (indicated by the single light-blue box in Fig. 5). ^{Reference BeAcares, Nazroo and Stafford44} However, as with the study of depression and anxiety in the Bangladeshi sample, ^{Reference Pickett, Shaw, Atkin, Kiernan and Wilkinson35} the Pakistani sample in this contrasting study of psychoses was mostly living in the most deprived category of neighbourhood.

Suicide and self-harm

One paper investigated the association between ethnic density and suicide ^{Reference Neeleman and Wessely50} and another on self-harm ^{Reference Neeleman, Wilson-Jones and Wessely51} (Fig. 6 and online Table DS4). Both studies were conducted in south London, and both included samples of African–Caribbean and Asian ancestry. Higher ‘non-White’ density was protective against suicide for the two ethnic groups combined, and similar associations were found for same-ethnic density, although these associations were not statistically significant. ^{Reference Neeleman and Wessely50} In the study of self-harm, higher ‘non-White’ density was associated with an increased risk of self-harm in one hospital catchment area but not another. ^{Reference Neeleman, Wilson-Jones and Wessely51} For those of African–Caribbean origin, higher same-ethnic density was also associated with self-harm in one hospital catchment area, but for Asian people higher same-ethnic density was protective in both areas.

FIG. 5 Tree map showing all independent analyses from studies of psychoses as boxes, labelled with reference number of paper of origin, scaled by sample size and shaded by result of ethnic density association (details in text).