The Problem with Categories: Towards A Relational Epidemiology

Beyond Identity

Epidemiologists and biostaticians need to think about they produce knowledge, not just how they describe it. If we stripped away traditional identity categories like race, gender, and age, epidemiology and biostatistics would look profoundly different. And, I think that's a good thing. It is time for us to look at shared experiences, (or exposures), as inclusion criterion and the define our populations, stratifications and categories, in ways that acknowledge real cultural differences and contextuality; not just what the government puts on your driver's license. The imposition of taxonomies predisposed identities, without consideration for the subjectivities of the individuals for whom the categories is set to represent. These representational strategies are then used to make generalizing statements, or snap shots, about people whose experiences are not embedded in the research. 

I'll give an example, then talk about the age-old Structure/Agency debate, and then present examples of what it would look like for public health research to go beyond the limitations of demography; and becomes truly inclusive of the diverse populations their research is designed  to serve. 

Do we truly need a "true sex"?

I'm thinking about Herculine Barbin, an intersex person whose memoirs, Herculine Barbin: Being the Recently Discovered Memoirs of a Nineteenth-century French Hermaphrodite, were brought to wider attention by Michel Foucault in the 1970s. Foucault's role in shaping modern thoughts on sexuality, especially gay culture, is unmatched. His examinations of power, identity, and discourse are foundational to Queer theory, LGBT studies, and the topicality of sexuality in general- but this post isn’t about him.

This post is about how Herculine Barbin’s story became a missing piece to the puzzle of epidemiological categories (or scientific identities), which I have long problematized. A random literary object, picked up by a militant philosopher, found its way onto my field of vision. Now, it’s all I can see…which is why I must write it out.

The singular use-mention of they was not conventional at this time, so Herculine used she/her pronouns.* There’s quite a lot to unpack just in that (women being the default for anomaly, or less than perfect examples of the ideal human form- a concept I wrote about in the previous post. She struggled, of course, with being identified in different ways, in different contexts, by different people, with different powers they imposed on her body. There was no freedom to identify yourself as such; just the sterility and inclusion and exclusion criterion developed by the magisterial minds of the moment.

“Oh! To live alone, always alone, in the midst of the crowd that surrounds me…” Herculine writes. There is weakened strength and a trickling of despair beading off this quote; its textuality speaks to the issue in question, as well as to the principles of holism. Isolation in the society and in the medicolegal world, meaning existential isolation, despair and nihilism in Barbin’s inner world; just as the variate definitions in the scientific and digital worlds reverberate the subjectivities of the cultural world in which we find ourselves now.

The question of a true sex, when there is enough evidence to make a broad definition as meaningless as a narrow one, is answerable only with a question of discourse as sites of power. Apart from our current preoccupation in all matters sexual, the story of this Herculine is important for it belies the rigidity of Western taxonomizing, especially where it concerns identity, be it race, class, gender, or sexual preference. That rigidity necessitates that all individuals be neatly pigeonholed and categorized according to a set of predetermined labels. So too in our epidemiological investigations we have concentrated on single-issue questions of identity, focusing singularly on gender or ethnicity, and have attempted to locate people from diverse communities and groups into a priori Western taxonomies: heterosexual/homosexual, male/female, elite/non-elite, etc.

Epidemiologists tend to concentrate on specific sets of issues that coalesce around topics like gender, age, or socioeconomic status, without interpolating other axes of identity, be they class, ethnicity, or sexual orientation for example, because this has been seen as too vast or complex a project. When I brought this issue up in my graduate internship with the Lead State Epidemiologist, she replied ‘everything cannot be studied at once’. This is where I see a fundamental disciplinary divide.

Critical public health scholarship is surely part of the project to open up the debate to all those vectors of difference by which individuals are named and subjectified. Pluralism makes possible the expansion of social ontology, a redefinition and redescription of experience from the perspectives of those who are more often simply objects of theory. Without adjustments to such sensitivities we run the risk of doing interpretive violence in representing people who do not neatly fit- or do not belong entirely- in the predetermined categories within which they are made to situate themselves; and by extension, imbricate themselves in the present-day struggle to be seen in the crowd.

Issues of identity have never been far away from epidemiology. Ethnic and racial categories, sex and gender (conflated early on), age, status, and other axes of differentiation have been core concepts underpinning epidemiological interpretation since Hippocrates. Recent approaches emphasize identity as multi-scalar, multidimensional, situational, and overlapping, constructed and negotiated by individuals in specific social contexts.

(I’ll note that this applies to population-level identification as well; and reframes much current thinking by envisioning community as constantly constructing—and being constructed by—constituent elements of demography, social interactions, cultural variations, complex information transfer, and manipulation of the environment in intra- and intergroup contexts in addition to the biological and ecological factors in development and throughout the course of life history.) It seems too big a project doesn’t it? …It is not.

Why It Matters

Theoretical approaches revolve around structure/agency debates: do we become ourselves from the pressures of society or can we shape our own identities through independent action.This dichotomous tension has important methodological implications. Researchers have to make a choice (bias), and decide how much weight to give to the influence of systemic structures versus individual decisions in their analysis. Overemphasizing structure might reduce individuals to passive actors at the whim of the collective. This is called victimization. Whereas an overemphasis on agency blames individuals for issues that are out of their control (systemic disparities). Both the ethics and politics of these perspective theories impact our basic interpretive mechanisms. How we understand meaning-making and social reproductive hinges on a balance between theories and the productive tensions produced.

So, in epidemiology, specifically, the question becomes: to what extent are health outcomes shaped by individual agency (choices) versus larger social, economic, and political forces (structure)?

If we lean towards agency, we ignore structural violence; and if we rest on structure, we endorse personal responsibility narratives. Structuralists engage in policy work, and systemic reforms; while agency-heavy approaches blame individuals for systemic issues.

Structure/Agency debates show up in data cross public health. Do we focus on environments (food deserts, for example), or education (behavior modification) in policy research and development. Statistical modeling questions if the focus is only on individual risk factors does the research miss the structural context? Our understanding of racial health disparities splits when culture and behavior are used as explanatory mechanisms rather than racism and white supremacy. And, of course, COVID-19 which primarily framed infections due to “poor individual choice” or crowded spaces (living, transit, social).

Bourdieu’s Theory of Practice & Giddens’ Structuration Theory are attempts to reconcile these disparate philosophies. Practical concepts like habitus (internalized dispositions shaped by social structures), field (social arenas with their own rules), and capital (resources people use to navigate these fields). He argued that while people are shaped by their social conditions, they also actively reproduce or transform them through practice. While Gidden’s claimed that structure and agency are mutually constitutive—structures shape human actions, but those actions also produce and reproduce the very structures.

The structure/agency debate is deeply connected to how epidemiological categories are constructed and used to measure and describe people; namely, those categories reflect assumptions about what matters in explaining health outcomes. 

Examples:

Category
Framed as Agency?
Framed as Structure?

Obesity
Lifestyle failure, lack of willpower
Food insecurity, neighborhood design

Race
Genetic difference (agency-biased misuse)
Structural racism, social inequality

Education level
Informed choice or self-improvement
Reflection of class-based access and opportunity

HIV-risk group
Risky sexual behavior
Effects of criminalization, poverty, stigma

These categories are not just data, they are sites of power and of control, imbricated in political, ethical, and conceptual decisions. What we measure reflects what we think matters (personal responsibility, social inequality). How we define categories can reinforce or challenge oppressive systems (race as biological or social, gender binaries, poverty as failure or systemic). Interpretation and action depend on framing (policy or education, reform or behavior change). The choices we make about this debate, in epidemiology, and for all of public health research, will have far-reaching effects on the outcomes of our work and its implications for building a healthier, equitable, and just society.

What would it look like to strip away demographic categories?

Instead of using broad identity categories, epidemiologist might rely more on:

Molecular/biological markers (e.g., viral load, genetic expression, immune signatures).

Socio-environmental exposures (housing quality, pollution exposure, food access, policing encounters)

Network structures (who interacts with whom, transmission pathways, kinship, migration)

Temporal patterns (life course transitions, timing of exposures, chronicity rather than chronological age).

This would shift the focus from “who are you?” to “what has happened to you, and in what contexts?" It would generate different statistical frameworks, as well. Instead of using traditional covariate or stratifying variables, epidemiologists might design models around exposure-outcome pathways as predictors. Moving beyond the prediscurive categories used now, and towards a greater use of latent variable models, (or clustering), patterns can emerge sui generis from the data. Instead of the just lumping two incommensurable problems together, a true focus on syndemics will reveal the interacting health problems shaped by structural conditions, rather than categorical disparities. For example, instead of "Black versus white HIV prevalence" the model might examine housing precarity, policing intensity, and STI co-infections as interacting forces. Now, we can start to see a true Epidemiologic "snap shot" of what the health is going in a given population. 

Reconceptualizing Health Inequities 

Traditional categories are often proxies of structural violence, used to avoid any critique of Capital and Empire. Without these representational strategies, researchers would have to build measures of racism, ageism, homophobia, and sexism as systems of exposure (discrimination indicines, spatial segregation, policy environments). Material conditions would need to be more than what GIS and GPS provide, to describe individuals environments based on income volatility, job insecurity, incarceration rates, and food deserts. Subjectivity (or lived experience measures) could include stress biomarkers, trauma histories, and individuated interpretations of well-being- and this has been done almost with the use of ACEs (Adverse Childhood Experiences).

To strip away race and gender risks erasing inequities only if not replaced with structural measures. When done correctly, this helps not harms minoritized communities and groups. It also reduces essentialism by spotlighting the environment rather than the body as the source of disparity. We don't need more categories of people, which is the current project in "woke" public health, (which is really not woke, and not progressive, it just wears a "woke" mask fit in while true radicals face the brunt of opposition for actually challenging the state of things). All kinds of new identity categories are being included in data collection tools and used in epi/biostats. 

To do without those categories, epidemiology and biostatistics could evolve into something more ecological, relation, and structural; and develop equally robust ways of captures the effect of culture, history, and agency. 

Here is a basic example of how can study something like HIV without saying who has HIV - which perpetuates stigma, as the category in text categorizes context- and instead, asking what systems, conditions, and exposures predict HIV transmission and care outcomes. 

Study Design

Unit of analysis: neighborhoods, networks, and exposures.
Variables collected:

Structural: housing precarity, incarceration exposure, policing encounters, insurance coverage, neighborhood segregation indices.
Biological: viral load, STI co-infections, PrEP uptake, immune profiles.
Network: sexual/needle-sharing network density, partner concurrency, migration/mobility.
Psychosocial: trauma histories, discrimination indices, stress biomarkers (cortisol, inflammatory markers).

Statistical Modeling
Rather than including “Black vs. white” as a predictor:

Models test how housing precarity + network density + incarceration exposure interact to shape HIV transmission risk.
Cluster analysis identifies emergent groups (e.g., “unstably housed, highly policed, high partner turnover” vs. “stably housed, insured, low network density”).
Outcomes compared across clusters, not racial/age/gender bins.

3. Findings
Instead of:

“Young Black MSM are disproportionately impacted by HIV.”

We report:

“Individuals experiencing housing instability, high exposure to policing, and dense sexual networks have HIV incidence rates 5× higher than those without these exposures, regardless of self-identified race, gender, or age.”

There is an agenda here just like with everything. There is no neutrality in science. By centering mechanisms rather than identities, we make structural violence visible as causal not as proxy; and we avoid reifying race as a biological risk factor. And policymakers won't need categories for priority setting and resource allocation, once when've built the structures of scientific revolution necessary to shift the paradigm towards mapping systems of harm: access, carcerality, housing, and hunger as the sources of risks. 

Here are some folks in public health that I found writing about this issue (although, not in the way that I am):

Nancy Krieger – social epidemiology, ecosocial theory

Michael Marmot – social gradient in health

Pierre Bourdieu – concepts like habitus and social capital

Ecosocial theory, fundamental cause theory, embodiment