How is ethnicity reported, described, and analysed in well being analysis in the United Kingdom? A bibliographical overview and center of attention team discussions with younger refugees | BMC Public Well being

Ethnicity refers to a multi-dimensional social construct encompassing language, country of origin, cultural heritage, nationality and more [1]. Since the 1991 Great Britain Census, ethnicity in the UK has been conceptualised as a subjective, self-defined construct, and commonly operationalised in research and administrative data under hierarchical ethnic categories with “Asian”, “Black”, “White”, “Mixed” and “Other” as the 5 high-level ethnic groups [2,3,4]. Reducing racial and ethnic health inequities has long been a priority for health policy in the UK [5]. However, ethnic health inequity remains prominent across medical disciplines, from stillbirth rates, access to mental health treatments, to cancer incidence and excess mortality from Coronavirus Disease 2019 (Covid-19)[6,7,8,9]. The National Healthcare Service Race and Health Observatory report [10] cites “lack of good-quality data and analysis” as the main barrier to addressing ethnic health inequities.

The ethnicity data gap reflects three interlinked issues [11]; i) the under-representation of people from minoritized ethnic groups in health research, ii) inconsistent and poor quality recording of ethnicity in health and administrative data and iii) the rigid, hierarchical ethnic categories used by researchers when analysing ethnicity that might not represent individuals’ and groups’ ethnic identity. In addition, how researchers analyse ethnicity data, and report (or omit) results of analyses of ethnicity, adds a further layer of complexity; 40% of randomised controlled trials funded by the National Institute for Health and Care Research did not report on ethnicity at all [12].

Under-representation of minoritized ethnic groups

The under-representation of minoritized ethnic groups is evidenced by human genetics research that uses a majority sample of European heritage, such as the UK Biobank (94% White), and is therefore not generalisable to UK populations (82% White) [13,14,15,16]. An alternative approach is to use linked population-based data sources that capture whole populations. There are an increasing number of health administrative data linkage studies aiming to improve diversity in the research samples, for example in the study of population vaccine uptake and variant phenotypes during the pandemic [17, 18]. However, linkage studies merely replicate data quality issues in the source datasets [19], which disproportionately affect minoritized groups [20, 21].

Quality of recording of ethnicity data

Issues of data quality may stem from the absence of guidelines on the collection of data on ethnicity and the breadth of collection approaches used (e.g. ethnicity being inferred from appearance, rated by healthcare workers based on proxy markers of ethnicity such as nationality, self-reported in fixed categories, or simply not collected). Selection biases may be amplified by approaches such as linkage, as minoritized ethnic populations are more likely to be missed and excluded in the process [22], and there are a variety of approaches to deal with inconsistent recording of ethnicity across linked data sources, or within a data source over time, despite this having important implications for results. For example, a recent study found the association between ethnicity, education attainment and neurodevelopmental disorder depends on which data source is used to primarily code ethnicity [23].

Ethnicity as a construct in research

Ethnicity is often analysed as a static construct, contradictory to its subjective, multidimensional and dynamic nature [24]. A large body of literature on longitudinal studies, health records and administrative data, demonstrates that people’s ethnic group identity changes over time and context, with more variation for non-white than white populations [21, 25, 26]. Within most population health studies, variation in recording of ethnicity is inadvertently treated as error. This has motivated substantive methodological work in harmonising these inconsistencies, for example using the most recent or most common (modal) ethnicities, or other weighted algorithms [27, 28]. These approaches remain insensitive to the dynamic nature of ethnic identity across time and contexts within local populations [11].

As the UK welcomes an increasingly diversifying population, there is a strong case for existing ethnic categories to be updated to reflect community identity. Whist recent census waves increased number of ethnic categories in the survey, from 9 categories in the first survey in 1991 to 19 categories in the 2021 census, existing administrative and health record systems have yet to catch up with such changes. Comparisons of UK Census data over the last 30 years identified a sizable increase in number of people identified as “mixed” and “other” groups [29]. The evidence base for these groups is often limited, as researchers struggle to understand what “mixed” and “other” mean [30]. The theoretical framework of a static ethnic identity faces conceptual and analytical challenges.

Meaningful analysis of ethnicity data

There is an urgent need for researchers to clarify how ethnicity is analysed in health research. In public health policy making, ethnic categories are rightfully used to foster comparisons to elucidate inequalities and prioritise resource distribution. However, the meaning of these categories is seldom made transparent, or communicated well to the public [11, 31]. What do we mean when we demonstrate that black women are 3.7 times more likely to die in pregnancy and childbirth than white women [32]? What do we mean when we model ethnicity as the cause of elevated risk for health outcomes? Theoretical frameworks for ethnicity in health have long been studied, from Nazroo’s [33] seminal work on how racism is fundamental to understanding ethnic health inequalities; Phelan and Link [34] describing how racism causes health inequalities independent of and in conjunction with social economic deprivations; Jongsma and colleagues [35] summarising the theoretical underpinning of ethnicity in relation to risks of psychotic disorders; to Bécares and colleagues [36] emphasis on racism acting at the structural, institutional, community and individual-levels leading to ethnic inequities in covid-19 vaccine hesitancy. It is less clear whether health researchers have been actively seeking to use and refine these theoretical models of ethnicity to help translate research into actionable health policies.

Reporting of ethnicity data

In 2021, the JAMA Network published updated guidance on reporting race and ethnicity in their journal, mandating authors to report how and from what data source ethnicity was classified, and what categories were used [37]. They encourage authors to explicitly theorise how ethnicity relates to health outcomes. Whilst some research has explored heterogeneity in reporting of race/ethnicity, such as the different ethnic categories used in the study of long-term conditions [38], most reviews on ethnicity and health still focus on describing health outcomes stratified by ethnicity and rarely report on how ethnicity is analysed. There is also limited involvement from the public in shaping how ethnicity is captured, and how research findings are interpreted to shape health policy [11].

The purpose of this bibliographical review is to examine the ways ethnicity is theorised, captured, reported, categorised and analysed in top-cited health research in the UK context. We also aimed to explore what factors researchers should consider when studying ethnicity in relation to health. We organised 3 focus group discussions with a group of young migrants and refugees from Coram Young Citizens to help us shape our study and interpret our findings.