7.3 Conducting intersectional sex-disaggregated and/or sex-specific analyses

 

Prior to analysing quantitative research data using an intersectional gender lens, data need to be disaggregated by relevant biological and social stratifiers. These may include sex, age, income status, disability, sexuality, geographical location, ethnicity, race, etc. This is true for research that is inter-categorical (e.g. analyses multiple social groups within and across categories) or intra-categorical (e.g. focuses on one social category at the intersection of multiple social identities in order to explain withingroup differences and larger social structures influencing their lives). In both approaches, the analysis focuses on the intersection of selected social stratifiers to understand how the stratifiers intersect to create different experiences of marginalization and discrimination, which in turn shape health outcomes related to infectious diseases.

Analysing quantitative research using an intersectional gender lens can involve multiple steps. These include conducting intersectional sex-disaggregated and/or sex-specific analyses and/or analysing outcomes against gender variables/gender equality indicators.

7.3.1 Conducting intersectional sex-disaggregated and/or sex-specific analyses

Below is a summary of the steps taken to conduct an intersectional sex-disaggregated and/or sex-specific analysis:

• Step 1: explore prevalence of disease between males and females (sex-disaggregated analysis).

• Step 2: explore prevalence of disease between and among different groups of males and females against different demographic variables (intersectional sex-disaggregated analysis).

• Step 3: explore within group differences among males and females using one demographic variable. For example, for all males who have TB, you disaggregate this data by age, education, geographical location, ethnicity/race, household earnings, etc. (intersectional sex-specific analysis).

• Step 4: explore within group differences among males and females using two demographic variables for those with a disease. This analysis is only conducted on the variables identified in step 1 that showed difference between groups. For example, if the highest prevalence of males who have TB are those with less than primary school, you take this sample and disaggregate it further by age, education, geographical location, ethnicity/race, household earnings, etc. (intersectional sex-specific analysis).

7.3.1.1 Step 1: Exploring prevalence of disease between males and females (sex-disaggregated analysis)

First, you will want to explore the prevalence of disease between males and females. Note that similar analyses can be conducted to explore risk factors/vulnerability for disease. Table 13 ►.

7.3.1.2 Step 2: Exploring prevalence of disease between and among different groups of males and females against different demographic variables (intersectional sex-disaggregated analysis)

Next, you will want to disaggregate your data even further by exploring differences between and among different groups of males and females related to disease prevalence. Different tables should be used for each demographic variable. (For example, there should be a different table related to age, education, geographical location, ethnicity/race, household earnings, etc.) Table 14 ►.

To facilitate the intersectional analysis, you will want to conduct sex-specific analyses (meaning data for males and females is analysed and presented separately).

The analysis below is sequential and includes two steps that build off the previous one to provide a more in-depth analysis of disease prevalence:

1. exploring within group differences among males and females using one demographic variable

2. exploring within group differences among males and females using two demographic variables

7.3.1.3 Step 3: Exploring differences in disease prevalence among males and females using one demographic variable

This section builds on the analysis above and disaggregates data further by relevant demographic variables. It includes information about males and females with a disease disaggregated by different demographic variables. You will want to conduct these analyses separately for males and females.

Table 15 ► includes data for those who have prevalence of the disease. Different tables should be used for each demographic variable (e.g. different tables related to age, education, geographical location, ethnicity/race, household earnings, etc.). For example, of all males who test positive for malaria, explore differences within this group in relation to age, education, geographical location, ethnicity/race, household earnings, etc. Table 16 ►.

7.3.1.4 Step 4: Exploring differences in disease prevalence among males and females using two demographic variables

This section conducts further analysis on the groups with the highest prevalence of disease. For example, if you have identified that males with 1-5 years of education have the highest prevalence of malaria, this section explores whether among these males there are further differences by age, geographical location, ethnicity/race, household earnings, etc. Table 17 ►.

Note: as you add variables, the N will decrease and the confidence interval (CI) is likely to increase. This can affect whether the sample is large enough to determine significance. While the sample sizes will be smaller, these analyses are worthwhile as they ensure that males and females are not treated as homogenous groups and allow for more tailored gender-responsive interventions. Table 18 ►.

To facilitate a gender analysis of the above sex-disaggregated and/or sex-specific data, the next section discusses how gender equality indicators and associated variables can be used. WHO & UNAIDS (2016) A tool for strengthening gender-sensitive national HIV and Sexual and Reproductive Health (SRH) monitoring and evaluation systems also describes how to conduct a gender analysis using gender equality indicators.

7.3.2 Using gender variables to conduct an intersectional gender analysis

Data cannot be disaggregated by gender in the same way data can be disaggregated by sex. Instead, relevant gender relations domains need to be included within data collection tools and interrogated separately; these are sometimes referred to as gender variables, and are used as proxies to understand gender relations.

Because it is difficult to ask about gender power relations directly, gender frameworks (See Module 3 ►) are used to break down the ways in which gender power relations manifest and develop proxies to analyse gender power relations against relevant health or other outcomes. While sex may be included as one variable within quantitative research, when using a gender lens, multiple gender variables will need to be included within data collection tools.

Examples of gender variables against a gender framework are presented in Table 19 ►.

Developing gender analysis questions through the creation of a gender analysis matrix ( Module 4 ►) will help you identify relevant gender variables, indicators and questions for data collection tools. As discussed in Module 3 ► and Module 4 ►, a gender framework can be used to develop gender analysis questions to be included within surveys and questionnaires.

Gender analysis questions might be related to access to different types of resources, distribution of labour and roles both within and outside the home, gender norms around what is or is not acceptable for a man or woman to do, and who holds decision-making power. The answers to these questions can then be interrogated against different social stratifiers and their intersections.

Table 20 ►presents gender analysis questions and their associated gender variables, gender equality indicator, data collection questions and source. Creating a similar table will assist you in the analysis of your quantitative data.

7.3.3 Classifying respondents to facilitate intersectional gender analysis

When interrogating the responses to the gender analysis questions against different social stratifiers, consider how the data is organized. Quantitative data is not always organized in a way that easily facilitates intersectional analysis (86,87). Data sets often collect data using one social stratifier at a time, i.e. sex or age, not sex and age together.

To explore the intersection of social stratifiers, you need to consider both sex and age together, for example, you need to know who a female and an adolescent is. The classification of respondents therefore becomes important, as well as including appropriate tracers within your data set to link participant classification.

Traditionally, demographic questions ask for a yes or no answer and are coded as 1 vs. 0 respectively (88). Detailed classification is needed, however, to create numerous categories. In order to ensure detailed classification, researchers can check for differences across social identity variables, and create an intersectional identity matrix that uniquely classifies each relevant subgroup (88).

For example, if you wanted to explore whether gender roles affect vulnerability to exposure to a vector-borne disease between males and females of different age categories, consider that differences exist between these groups. Four or more variables need to be distinguished and classified, as opposed to having separate variables for age and sex.

Table 21 ► below present four variables in which individuals are classified, combining age and sex into one variable. This ensures that throughout the analysis individual experiences are not lost, and analyses are more robust. Such an approach can be problematic when working with a limited sample size.

According to Rouhani (2014), a challenge remains in balancing number of categories and maintaining adequate statistical power. They suggest that researchers either increase the sample size to improve statistical power to account for the multiple categories or increasing the conventional alpha level from p <0.05 to a higher cut off, such as p<0.10 in the analysis.

In Table 21 ►, in order to conduct an intersectional gender analysis, the variables adolescent male, young adult male, adolescent female and young adult female can be analysed against the responses to the gender analysis questions to look for differentiation across the different groups

7.3.4 Analysing quantitative data through an intersectional gender lens

Within Intersectionality-informed Quantitative Research: A Primer, Rouhani (2014) discusses multiple approaches to intersectional analysis, including additive (unitary) and multiplicative.

According to Rouhani, traditional quantitative methods utilize an additive approach to examine individual effects of various factors on a given outcome when controlling for other variables.

Intersectionality-informed analysis uses an additive approach as initial ‘baseline’, upon which further analyses are applied using multiplicativity (e.g. regression coefficient) to account for effects of intersecting categories on health or social outcomes. This enables researchers “to determine whether two-way, three-way or four-way statistical interactions (i.e. intersections) between axes of inequity contribute to explaining variability in a given outcome above and beyond the additive approach” (Rouhani, 2014: 9).

For additional explanations of these approaches and examples of equations and analysis techniques, see Rouhani (2014).

7.3.5 Interpreting quantitative data through an intersectional gender lens

An intersectional gender analysis within quantitative research also comes into effect during the interpretation of results.

When interpreting data from an intersectional gender lens, the researcher needs to put the data and results into context, particularly in relation to the historical and contemporary structuring of inequalities within the wider society and among individuals in study.

During analysis and interpretation of the results, understanding the context allows researchers to better interpret and make sense of data, and understand the drivers and mechanisms of inequity and what might be done about it (87). Results should be interpreted and understood against differential gender norms or roles, in addition to other social and structural inequities, including ageism, classism or racism.

Findings from a study exploring the prevalence and risk factors of schistosomiasis among Hausa communities in Kano State, Nigeria were presented in Module 3 ►.

• The study found that the prevalence of schistosomiasis was much higher among males (20.6%) than females (13.3%) in the sample (53).

• Disaggregation by age showed that prevalence was highest among the 11-20 age group (27.4%), followed by the 21-20 age group (14.4%).

While these stratifiers were explored separately, and one can surmise that prevalence is highest among males aged 11-20, an intersectional analysis would combine these categories to explore prevalence among males and females within different age groups, which would potentially tell a different story.

Applying an intersectional gender lens to this would help us understand why prevalence is highest among the 11-20 age group and among men/boys. In such settings, for example, adolescent boys or young adults often have much more freedom to swim in bodies of water, either due to having fewer domestic responsibilities or gender norms. This places them at higher risk of being exposed to schistosomiasis than women/girls of the same age.

While the results can be interpreted against different gender relations domains, including gender analysis questions related to gender roles and norms with the study tools would provide specific and robust evidence about the role of gender power relations in exposure to schistosomiasis. Follow-up qualitative studies would enable a more in-depth exploration into the role of gender relations and their intersection with different social stratifiers.