Background The incidence of femicide, the murder of a woman for gender-based reasons, has risen significantly in recent years in Mexico, going from 427 reported victims in 2015 to 797 in 2024. Although individual-level risk factors have been studied, we aimed to study neighborhood-level factors that might be more amenable to intervention. Our study delved into the impact of municipality-level violence in Mexico on the probability of occurrence of domestic femicides at the municipality level in 2024. Methods We employed the Causal Roadmap framework for causal and statistical inference. Our outcome was the occurrence of at least one domestic femicide in the municipality, with exposure defined as above-median male homicide rates. While unmeasured confounding precluded definitive causal inference, we estimated the adjusted risk ratio and adjusted risk difference (aRR, aRD) comparing municipalities with above-median versus below-median male homicide rates, adjusting for measured municipality-level confounders. We used targeted maximum likelihood estimation (TMLE) with Super Learner to minimize reliance on parametric modeling assumptions. Results Municipalities with high male homicide rates (above the median) had 4.71 times the risk (95% CI: 3.55-6.24, p<0.001) of experiencing at least one domestic femicide compared to municipalities with low male homicide rates, after adjusting for socioeconomic factors using targeted maximum likelihood estimation. Results were robust across three causal inference methods: TMLE (aRR=4.75, 95% CI: 3.55-6.24), IPW (aRR=4.75, 95% CI: 3.58-6.29), and G-computation (aRR=4.33, 95% CI: 3.32-5.65). Conclusions Using the Causal Roadmap framework, we found a strong and consistent association between municipality-level male homicide rates and domestic femicide occurrence in Mexico. These findings suggest that interventions targeting community-level violence may have downstream effects on reducing domestic femicides, offering potential strategies for comprehensive violence prevention in Mexico.
Armenta et al. (Mon,) studied this question.