In a consensus process, the INGER research consortium designed questionnaire modules that largely meet the requirements of the sex/gender concept which was developed in a parallel process. This process was based on a synopsis of previous sex- and gender-related approaches and their operationalization examples (26) as well as on further extensive research on possible survey tools on sex and gender. Fundamental to the development was a comprehensive decision-making process between the experts of the scientific consortium including the scientific advisory board. For the implementation of survey tools, an iterative discussion procedure was initiated, inspired by the nominal group technique (30). Initially, possible survey tools and their impact and historical context were presented in an internal gender-theoretical discussion paper. A second discussion paper from a gender-theoretical perspective weighed up the advantages and disadvantages of the different instruments. The consortium members also contributed further instruments to the discussion process and voted on all instruments in a consensus process during several discussion rounds in 2018. The scientific advisory board of INGER debated a first version of the questionnaire modules in spring 2018. The piloting took place in autumn 2018.

The final questionnaire was developed in German, but the individual questions and response categories are translated into English here. The individual sex/gender self-concept was developed for general application in various fields such as epidemiological studies or human biomonitoring research. The same is true for the topics that were considered important for contributing to explain structural sex/gender relations. However, items selected for each topic may vary depending on the target population.

The inner circle of the graphic of INGER's multidimensional sex/gender concept, the individual sex/gender self-concept, has five dimensions: sex phenotype at birth/sex assigned at birth, current sex phenotype, sex/gender identity, internalized sex/gender roles, externalized sex/gender expression (26).

In our society, sex is assigned at birth based on the observable phenotype and influences all other dimensions without determining them. We assessed sex/gender as a two-step approach enabling the investigation of external attribution (sex assigned at birth) and self-categorization (current sex/gender identity). This approach assesses more than two forms of sex/gender (e.g., female, male, inter^*, trans^*) and that a sex/gendered body at birth does not necessarily match a corresponding sex/gendered identity at the time of the study in adulthood. The category “sex assigned at birth” enables comparability with other studies that measure sex/gender with a binary category (male/female) by including the legal sex/gender status. In addition, self-categorization allows space for individual sex/gender identity conceptions. The two-step approach has been evaluated repeatedly in recent years (17, 34, 46) and is now increasingly recommended and used (33, 47–49), allowing for a more accurate mapping of sex/gender minority positions. It was pointed out, that despite slightly different categories, all two-step approaches capture three dimensions of sex/gender: sex assigned at birth, sex/gender identity and trans status (17). Response options used in INGER included “I do not want to classify as any sex/gender category”, which considers individuals who do not identify as any of the other given response categories. However, to overcome the binary construct to more appropriately reflect sex/gender diversity an explicit classification “non-binary” should be added in future studies. An interesting new instrument to map and visualize diverse sex/gender identities is a diagram published by Beischel et al. (23). In addition to cis and trans, it also classifies a category of allogender (neither cis nor trans) and distinguishes between binary and non-binary identities of cis, trans and allogender persons. However, a 12-min instructional video must be watched before use, during which 50% of participants dropped out in the corresponding feasibility study (23). In addition, 19% of the feedback on the instrument from the fully participating participants was negative (22). Therefore, this instrument seems to be inappropriate for the use in quantitative epidemiological studies. For our questionnaire, it would be useful to additionally explain the terms cis, trans^*, inter^*, and non-binary to all study participants.

The use of the two-step approach in data collection proves feasible in the KORA cohort. Missing rate was extraordinary low and comparable to other sex/gender- and not directly sex/gender-related variables in our survey. As expected, prevalence of non-cisgender persons, i.e., persons with a current sex/gender identity not matching their sex assigned at birth, was very low in our study (0.45%). This result based on the KORA cohort should not lead to a general conclusion that the two-step approach is superfluous in epidemiological studies. The observed prevalence seems to reflect the reality of the distribution of non-cisgender persons, as it is comparable to or even higher as the prevalence estimates of other only rare, population-based surveys, that were predominantly conducted in the United States (50–52).

To measure internalized sex/gender roles, we used the TMF scale, an instrument from psychology—new at the time of questionnaire development—which considers femininity and masculinity as higher order constructs (35). Sex/gender roles are usually measured in psychological studies by scales for personality attributes, behaviors, interests, and attitudes with the Bem sex role inventory as one prominent example (12, 53). The TMF scale allows self-categorization without assigning stereotyping and culture-specific concrete traits and is thus also internationally valid in different cultural areas. In addition, it easily separates the constructs of masculinity and femininity. As sex/gender roles are constantly changing, trait-based scales can no longer reliably assign traits to different sex/genders (35, 54). While this instrument leaves open which concrete traits a respondent associates with femininity and masculinity, it may reveal an internalized sense of femininity and masculinity as a characteristic of sex/gender. However, the TMF scale assumes binarity which we basically wanted to overcome. Nevertheless, it helps to investigate whether measurable sex/gender health differences can be related to the self-concept rather than the sex assigned at birth (e.g., risk taking behavior—which is assumed to be a masculine trait—in females with a low feminine sex/gender role concept (55).

To further assess a person's gender role attitude, we applied items of the proven instrument of the ISSP, suitable in content for the middle-aged to older persons of the KORA population (36, 37). However, some items of the instrument are not entirely clear in their construct validity: e.g., “One parent can bring up a child as well as two parents together” and “A same sex couple can bring up a child as well as a male-female couple.” Both statements could also be rejected when structural barriers impair equal opportunities. Nevertheless, this instrument allows us to look at the influence of internalized gender roles on health. In particular, the discrepancies between sex/gender role attitudes and actual gender role behavior could have a negative effect on the health-related quality of life if these discrepancies lead to subliminal stress and in a long run to an increased allostatic load. Cumulative allostatic load has been discussed as one major biological pathway linking social inequalities with ill health (56, 57). However, effects of gender role attitudes and behavior on health status have not yet been the focus of health research and should therefore be investigated more intensively in the future.

In general, the participants in our survey tended to have a more egalitarian attitude. But we observed greater proportions of traditional attitudes in the older age groups, which suggests that gender role attitudes adjust over time with societal changes. With respect to the feasibility, missing rates was low except for the question if being a househusband is just as fulfilling as working for pay. Of the participants who did not answer this question, 74% had a female sex/gender identity who may have found it difficult to answer this question from a rather unusual househusband's perspective.

Externalized sex/gender expression as perceived by others was assessed based on a short self-report measure for use in the study of health disparities (39). Combined with sex assigned at birth it can be used to evaluate sex/gender non-conformity (18, 38, 39, 45) which may be a source of detrimental health effects. Individuals who are perceived by society as sex/gender non-conforming and thus deviate from an expected appearance may suffer from long-lasting psychological distress and somatic complaints because of discrimination (58, 59). Accordingly, non-conforming adolescents have been shown to have a lower health-related quality of life (38) or lower self-reported health (18). In our analysis, we evaluated the non-conformity based on the external attribution as well as on the internalized sex/gender roles. In line with Hart et al. (18), we found in the KORA population that less than one-third of participants selected the highest category (very masculine or very feminine) for each item that corresponded to their sex assigned at birth. Additionally, Hart et al. (18) found that only people for whom the self-rated non-conformity did not match the assumed appraisal by others reported poorer self-related health. When we classified our participants either as conforming or as non-conforming, 1% assumed to be viewed as non-conforming by others but had an own perception of conformity and therefore may experience distress and on the long-run health impairments. This underscores the benefit of moving beyond a binary categorization, that does not capture the different degrees within femininity and masculinity. It should be noted that the corresponding items had slightly higher missing rates (4.7–6.9%) in our survey. Although unsolicited comments are not representative for the entire study population, they indicated that the terms “masculine” and “feminine” were sometimes questioned or not understood. When using these scales in the future, the terms should be explained. Another reason could also be the use of a bipolar scale from “very masculine” to “very feminine”. The assessment of the extent of masculinity and femininity may vary in different situations and may not necessarily be mutually exclusive (39). Therefore, we suggest changing the response scales in upcoming studies. Instead of a bipolar scale, two separate scales should be used, as it was also applied recently by Hart et al. (18). One scale should range from “not at all feminine” to “very feminine” and one from “not at all masculine” to “very masculine”. Thus, gender as a bipolar construct is replaced by a more contemporary design that allows for femininity and masculinity at the same time. It also better accounts for non-binary individuals who do not identify as either feminine or masculine, who might answer “not at all” to both scales.

In order to adequately map hierarchized social relations between different sex/gender groups, social context factors can add information on structural living conditions leading to social inequalities between diverse population groups. For health, the perceived burden of disadvantage is particularly important (60). Therefore, we used an instrument to measure experiences of discrimination due to several reasons. In general, the study participants felt rather not or not at all discriminated against. As is known from anti-discrimination research in the social sciences, experiences of discrimination, however, do not adequately reflect the actual reality of inequalities. Often, affected persons may not directly recognize disadvantages and individual feelings of disadvantage can be less pronounced than the disadvantage of one's own social group (61, 62). This is in line with the low level of feelings of disadvantage because of sex/gender in the comparatively affluent KORA research population, despite the evidence that more generally at the societal level women are discriminated against according to objective measures such as the Gender Pay Gap.

The INGER assessment of discrimination allows for the consideration of subjective experiences that may have a direct effect on health or may alter vulnerability to environmental exposures. In order to assess the whole spectrum of discrimination and to overcome the data gap regarding marginalized groups in health science, their representation in processes within the health care system (e.g., diagnosis and treatment frequency) should also be examined. However, for the German-speaking area, a consensus on appropriate instruments for surveying identities of marginalized groups (e.g., certain sex/gender identities, cultural/ethnic identity, and sexual orientation) is still missing, but would certainly be necessary. Measuring external attribution is particularly useful for surveying discrimination, which is why we included the question “Have you ever been asked in Germany whether you or your parents were born abroad?”, an instrument on self-assessed othering processes (41). It shows whether a respondent is segregated from the collective of the community and allows for asking the presumed individual characteristics as cause for the question. Such measurement tools are still missing for other reasons of discrimination. Respondents in our study reported little experience with othering processes, which may be because only persons with German citizenship were invited to take part in the four KORA surveys between 1984 and 2000.

One of the greatest inequalities in sex/gender relations at least in the society of Germany is the gender care gap [Second Equality Report of the Federal Government, (63)]. Correspondingly, in our study we observed that women more often had the sole responsibility for childcare or care of disabled people in the personal environment, housework, cooking and shopping. The gender care gap can lead to unequal physiological and psychological distress (64, 65). In addition, this distribution of care and household work has an unequal impact on people with different socio-economic positions. For example, household and care work in high social classes is mainly delegated as paid work to (immigrant) women from low social classes and only people from higher economic positions can afford to delegate household and care work (66, 67).

An adequate analytical integration of sex/gender relations can only take place if structural living conditions such as discrimination, care and household work are included in health studies. The INGER study has operationalized essential aspects here for the KORA population, but further items should be included or developed that capture sex/gender relations more in depth and allow for further differentiation between marginalized groups depending on the specific population group being studied (e.g., job roles, access to public/private childcare, residence status, access to medical care, geodata on crime rates/social housing in the respondents' neighborhood, perceived safety, body sovereignty).

A major strength of our comprehensive sex/gender questionnaire is that we operationalized all items according to a theory-based sex/gender concept that reflects the current state of knowledge on integrating sex/gender in health research. In addition, the response rate was very high, emphasizing the feasibility at least in middle-aged and older population-based cohorts. Furthermore, the operationalization of the sex/gender concept can be applied particularly well to already existing populations. It is possible to draw on already existing data fitting to the concept (especially intersectionality-related social categories), which has a positive effect on the required length of the questionnaire. In particular, the operationalization of the concept was implemented in such a way that all groups, including marginalized ones, are adequately taken into account in data collection. However, existing cohorts might not reflect the diversity of the general population in terms of immigrant groups, low-income groups or people with marginalized gender and sexuality in sufficient case numbers to be able to perform quantitative analyses e.g., on their structural sex/gender relations. This was also the case for our study population with very low or even no persons belonging to marginalized sex/gender identity groups. Therefore, when planning a new study with new recruitment of study participants and depending on the specific research question, an appropriate group-specific power calculation should be made and strategies for contacting and motivating all relevant population groups should be determined (68). Lastly, we collected data on characteristics related to social inequalities separately for relevant categories (e.g., separate questions on sex/gender, socio-economic position) to allow for the application of a variety of statistical methods to analyze the complex interactions of social categories.

Nevertheless, there are also a few limitations of our study. We were not able to include a measurement for the biological sex in our study. The question on sex assigned at birth, which aims to describe an unchangeable state of sex, corresponds to a cultural attribution practice in which an external appearance of a body is categorized according to social standards. Thus, it does not describe all of the physiological details and variability within sex groups and is therefore not meaningful enough as the true biological sex. For a more differentiated assessment of sex, biological measurements (e.g.„ of chromosomes, hormone concentrations, metabolic processes) would be necessary depending on the research question (12), which could not be implemented within the framework of our data collection by self-administered questionnaires.

The questionnaire instruments originate from a European and North American context. Therefore, our operationalization might not be applicable for e.g., indigenous concepts of sex/gender (69) as well as sex/gender relations from South America, Asia or Africa.

To address the presumably low prevalence of marginalized groups, we reemphasise the importance of planning sufficient case numbers when establishing a new study population. This will also show whether the operationalization leads to the same results in terms of response and missing rates as in our middle-aged to older cohort. Furthermore, on the basis of our research interest in health effects of environmental exposures, we collected data on perceived discrimination due to sexual orientation or ethnic/cultural identity as possible causes of chronic stress, but did not also include items on self-assessment of sexual orientation and of ethnic/cultural identity in our questionnaire. Analyses stratified further by categories of sexual orientation or ethnic/cultural identity may be relevant in sex/gender related health (70). However, data collection on sexual orientation has been rather uncommon up to now in German epidemiological studies (71) and community-based developed instruments for both areas are currently still lacking in Germany. In the context of research on discrimination processes, first suggestions to measure sexual orientation (72) and cultural/ethnic identities (47, 61) in German surveys have been made. Involving community members as active and equal participants in the questionnaire development process would ensure that relevant categories are captured for all, including marginalized groups (73). Such a community-based participatory approach would help to develop more comprehensive instruments for data collection, but this was clearly outside the scope of our research aim in INGER.

Intersectionality focuses on the close entanglement of social categories and thus interactions of social inequalities (74, 75). The often explorative—analysis of complex interactions of a wide range of social dimensions calls for new statistical approaches (24, 76). In environmental health research in particular, there is a need for methods that can be used to analyze in an intersectionality-informed way whether sex/gender, captured by multiple sex/gender-related variables, has an impact on an environmental exposure or on the association between environmental exposure and a health outcome (8, 77). By combining manifold newly collected data based on the multidimensional INGER sex/gender concept with already available data for an intersectional perspective, the operationalization for the KORA population provides an innovative data base for such analyses. For example, within the INGER project we were recently able to show that decision tree analyses are useful for exploring the relevance of multiple sex/gender dimensions, further social categories and their interactions for the exposure to green spaces (78).