Abstract Background Sex and gender both affect health outcomes, often in complex ways that intertwine biological and social influences. While researchers have criticized the conflation of sex and gender in quantitative studies, it remains a challenge to analytically disentangle them. We investigated an approach to conceptualize sex and gender as interrelated constructs embedded within a specific social context and estimate their direct effects on health outcomes simultaneously. Methods To analyze migraine and stroke incidence from 2011 until 2022, we used longitudinal data from the representative German Socio-Economic Panel (SOEP) within a causal framework. We applied a directed acyclic graph (DAG) to formalize hypothesized effects of sex and gender. Gender was modeled as an unobserved latent variable, influenced by sex assigned at birth and indicated by a set of gender-related variables. We codified the causal model as a structural equation model (SEM), enabling the joint estimation of the latent gender construct and the direct effects of both sex and gender on the health outcomes. We tested prior hypotheses in the SEM and explored additional relationships using causal discovery techniques. Results Migraine incidence was 3.1 times higher among participants of female sex (6.2% vs. 2.0%). Sex had the strongest effect on reporting a new migraine diagnosis ( β = 0.044, 95% CI 0.033, 0.056, p < .001), while gender showed no direct effect. Stroke incidence was lower among female respondents (1.2% vs. 1.7%), with sex showing a statistically significant negative effect, indicating that slightly fewer female persons reported a new stroke diagnosis ( β = -0.010, 95% CI -0.017, -0.004, p < .01). In contrast, gender showed a small, but significant positive effect ( β = 0.006, 95% CI 0.002, 0.009, p < .01), suggesting that gender-related characteristics that were more frequently reported by female individuals had an effect on stroke incidence. Conclusions Both sex and gender differentially affected stroke, whereas only sex showed a direct effect on migraine. We demonstrated that biological and social dimensions of sex and gender can be systematically addressed within the same model including gender as an unobserved latent variable, while remaining attentive to contextual complexity.
Kern et al. (Mon,) studied this question.