The scoping review followed the methodology described by Arksey and O’Malley in 2005 [25] and further refined by Levac in 2010 [26]. They suggest and elaborate on five central methodological steps: 1) Identifying the research question, 2) Identifying relevant studies, 3) Selecting studies, 4) Charting the data, and 5) Collating, summarizing, and reporting results, as well as a sixth optional step: 6) Consultation. Furthermore, we embraced the definition by Colquhoun and colleagues from 2014, which points out that a scoping review “addresses an exploratory research question aimed at mapping key concepts, types of evidence, and gaps in research related to a defined area or field by systematically searching, selecting, and synthesizing existing knowledge”.
Identifying the research question
Through an iterative process of trial and error, discussions and consultations the scope of the review was formulated in terms of aim, research questions and eligibility criteria. For example, several dimensions of inequality were considered using the PROGRESS-Plus framework [27]. The dimensions finally included were: socioeconomic position (education, income, occupational class, etc.), gender, race or ethnicity, sexual orientation and religion. Age and disability were at first also considered but finally excluded. Disability was excluded mostly due to the difficulties found with respect to the screening process in which studies with mental disability as an outcome had to be discriminated from studies with mental disability as an exposure. We came across other difficulties with age as it is routinely included in analytical models as a covariate and thus difficult to both identify from abstract and interpret. The scope was also limited to high-income settings in order to avoid too much heterogeneity related to for example norms about LGBT-persons, welfare systems and division of labor between women and men.
In addition to the overall aim of mapping, describing and analyzing intersectional inequalities in mental health, four research questions were formulated in order to provide a roadmap for the subsequent stages:
-
1.
Which are the intersectional social positions studied?
-
2.
How has intersectional inequality been operationalized?
-
3.
Which intersectional inequalities in mental health emerge in:
-
4.
Which explanatory factors have been analysed with respect to intersectional inequalities?
While the aim guided the development of the search strategy and the selection of studies, the research questions were applied to explore the finally included studies.
Identifying relevant studies
The search strategy was developed together with an information specialist. Due to the breadth of the search, and thus the large number of studies identified, the search was limited to two electronic databases: PsychInfo (American Psycholological Association, APA database) and the National Library of Medicine’s PubMed (including Medline). No language restrictions were applied to the search. Articles in foreign languages provided the title and abstract in English and they could undergo the screening process without translation. The search strategy included two complementing, but still overlapping, searches with three search blocks each. The first search focused on inequalities linked to socioeconomic position, and the second on inequalities linked to well-established dimensions of discrimination: gender, race/ethnicity, religion and sexual orientation. Both searches included a block defining the outcome, i.e. aspects of mental ill-health. The full electronic search strings, which comprised both Mesh terms and free terms, is provided as Additional file 1. The original period applied to the search was 1st January 1997 to 26th January 2017. A second search for papers published between 27th January 2017 and 25th January 2019 was added using the same search terms. The full search process is illustrated in the flow diagram (Fig. 1).
Screening and selecting studies
Primary research studies from high-income settings, and with a majority of the participants over 18 years, were included. Thus, editorials, letters and reviews were excluded. For eligibility, the study also had to analyze and report inequality defined by intersections of socioeconomic position (education, income, occupational class, etc.), gender, race/ethnicity, sexual orientation or religion. Finally, the mental health outcome could either be self-reported symptoms assessed through validated scales, or mental health disorders assessed through diagnostic interviews.
The search result was imported to the reference management software EndNote. A specific layout and screening procedure described by Bramer et al. in 2017 was used for screening efficiency [28]. Two researchers (AM and NFT) assessed the literature for eligibility independently from each other in two stages: first, title and abstract screening and second, full text screening. An initial meeting was held after screening title and abstracts of a few hundred studies. The purpose of the meeting was to clarify any questions regarding the interpretation of the eligibility criteria in order to ensure inter-rater reliability. A similar meeting took place in the process of screening the studies selected for full text reading. Any disagreement between reviewers in the full text screening process was resolved with a third party (PEG). The result from the second search was screened by NFT and the studies included for full text screening by both NFT and AM.
Charting the data
As a tool for systematic data extraction, a chart was developed by the authors (AM, PEG and NFT). The data chart initially included columns for the basic characteristics of the studies such as publication year and author. It was further developed to include information on study composition by columns for study context, population size and age, and outcome measure. Finally, columns for data specifically corresponding to the research questions were included in the chart, i.e. the intersections investigated, how intersectionality was operationalized, the direction of any statistically significant intersectional inequality and how it was estimated in terms of absolute and relative measures (mean difference or odds ratio (OR)) as well as potential explanatory factors and their explanatory value. The data was chartered by NFT and cross-checked by the other authors (AM and PEG).
For the extraction of data on intersectional methods and results, we adopted the terms used by Jackson et al. as a terminological framework [29]. According to this, intersectional inequality in health between the two doubly disadvantaged and the doubly privileged position is called the joint intersectional disparity and the inequality between the middle groups and the doubly advantaged position as the referent disparities. The difference in health between the joint intersectional disparity and the sum of the two referent disparities equates to the excess intersectional disparity, which corresponds to the interaction effect between the two intersecting positions. If the excess intersectional inequality was found to be positive, it was labelled as having a synergistic effect, and if it was negative, it was labeled an antagonistic effect. The original approach by Jackson et al. use these terms solely for absolute inequalities (e.g. mean difference), but for the purpose of this review we also apply the terms to relative inequalities (e.g. OR).
Assessing the risk of bias
We decided to assess and report the risk of bias in the individual studies in a separate chart. Since published and validated risk of bias tools are not designed for assessing studies about inequalities, even less about intersectional inequalities, an assessment guide was developed. Social inequalities in health is a phenomenon that can be constructed as a statement about the population distribution, and there is no evident way to handle confounding vs. mediating factors in relation to exposure and outcome. We therefore focused less on items concerning a causal relationship. The following five domains of quality and risk of bias were critically assessed:
-
1)
Study design
The study is preferably primarily designed to analyze intersectional inequalities (e.g. stated as aim or objective and not included as post hoc analysis).
-
2)
Data collection
The study population is randomly selected and preferably representable of the general population. The collection of data on background characteristics such as income and social class is preferably based on registers, and personal characteristics such as sexual orientation is based on validated questions or questionnaires.
-
3)
Outcome measures
Validated scales or diagnostic tools should be used to measure the outcome.
-
4)
Statistical methods
Preferably statistical methods yielding results corresponding to the intersectional inequalities described in [29] are applied.
-
5)
Reporting of intersectional inequalities
Preferably, all results, not only significant estimates, are reported.
Two authors (NFT and AM) assessed the included studies independently from each other, and a third party (PEG) resolved disagreements, with the exception that the authors did not assess any of their own articles. For each item fulfilled, one point was rewarded, thus the maximum quality ranking per study was five points. The assessment was done to provide a general overview of the quality of the results extracted into this review and not as part of the eligibility process.
Collating, summarizing, and reporting the results
The charted data was primarily collected and combined according to the intersections identified, for example, the intersection of race/ethnicity and gender. The literature was described according to its basic characteristics, the number of studies covering the specific intersection, and the way intersectional inequalities in health was operationalized according to the terminology of Jackson et al. [29]. For each intersection, we summarized the direction of a statistically significant intersectional inequality, and how it was estimated. The reporting of these summaries as well as data about explanatory factors were structured according to outcome measure and type of intersectional inequality. In this way, data was gradually configured into a descriptive narrative of the literature about intersectional inequalities in mental health.
Consultation
The review was commissioned, as part of a governmental assignment, by the Public Health Agency of Sweden, which is the national body responsible for monitoring health inequalities in the Swedish population. As such, the progress and results were shared and discussed continuously throughout the work with representatives of the agency. This kind of stakeholder involvement was regarded as important in order to increase the uptake of results, including the intersectional analytical approaches described and discussed in the article which could be used to refine national and regional monitoring. Consultations were done by providing preliminary drafts for feedback, and through dialogue according to pre-specified dates.