The ecological study by Keown et al ^{Reference Keown, McBride, Twigg, Crepaz-Keay, Cyhlarova and Parsons1} is undoubtedly of value, both methodologically and in relation to the further exposition as to how ‘structural and social issues can shape mental health’, as Burns and Rugkåsa ^{Reference Burns and Rugkåsa2} (p. 97) note in their related editorial. However, some clarification of the authors' use and operationalisation of the term ‘ethnic density’ is required to more fully understand the study's implications and limitations.

The study documents ‘a positive association between ethnicity and compulsory in-patient treatment’ in urban areas (p. 158), but as the denominator of population analysis is relatively large (divided by primary care trusts (PCTs) with an average population of 350 000), it is unclear whether ‘ethnic density’ is defined in their study according to the overall prevalence of different ethnic groups within these relatively large unit PCT populations under study, or whether smaller and more relevant unit neighbourhood-level measures of ethic density have been used.

An important earlier study using such neighbourhood-level measures, by Das-Munshi et al, ^{Reference Das-Munshi, Bécares, Boydell, Dewey, Morgan and Stansfeld3} demonstrated that ‘people resident in neighbourhoods of higher own-group density experience “buffering” effects from the social risk factors for psychosis’ (p. 282). As psychotic presentations are more likely to result in compulsory admission, Das-Munshi et al's findings would be expected to predict, when controlling for other variables highlighted by Keown et al – in particular, age and deprivation indices – that higher ethnic density, through ‘buffering effects’, would lead to lower levels of compulsory admission. Although it is possible that the findings of Das-Munshi et al and Keown et al are therefore in contradiction, it seems more likely that the Keown et al study did not measure ethnic density at the more relevant neighbourhood level in which buffering effects are manifest, and therefore that their measure of ‘ethnic density’ is less meaningful.

Ecological studies, by definition, attempt to attend to these more proximal influences on the immediate living environment. ^{Reference Jadhav, Jain, Kannuri, Bayetti and Barua4} Although the data-set used by Keown et al no doubt precluded this, the contingent limitations of such data, if this was the case, are therefore important to further acknowledge. Neighbourhood-level ethnic density data would also be needed to confirm the significance of Keown et al's unexpected finding of a lack of association between ethnicity and compulsion in rural areas, where genuine neighbourhood-level ethnic density might be expected to be low, at least in some areas. Nonetheless, Keown et al's study alerts us to the importance of attending to both social and cultural factors influencing the genesis, precipitation and maintenance of mental illness, including psychosis, which may be variously protective or risk-amplifying, and which interact in complex – sometimes counterintuitive – ways, influencing prognosis, ^{Reference Rodger and Steel5} hospital admission and compulsion.