Antidepressant discontinuation in France from an intersectional perspective: a discrete-time survival analysis within the MAIHDA framework

Antidepressant (AD) discontinuation before remission is common in depression care, despite guidelines recommending continuation for six months after remission to prevent relapse. Evidence suggests that socially disadvantaged groups are more likely to discontinue treatment despite higher depression severity, which may contribute to widening inequalities in mental health care. Most studies focus on single social characteristics rather than considering complex intersecting social identities. This study examines AD discontinuation over the first year of treatment across intersectional social strata. We analysed data from AD users in the French CONSTANCES cohort (N = 44,268; 64% women) linked to the national drug reimbursement register (2007–2021). We assessed AD discontinuation using discrete time survival analysis within a Multilevel Analysis of Individual Heterogeneity and Discriminatory Accuracy (MAIHDA) framework across 36 intersectional social strata based on unique combinations of gender, education, generation, and migration background. Social inequities in AD discontinuation between intersectional strata slightly increased over time. Men or women intersecting lower education, younger generations, and migrant backgrounds had a 44% predicted probability of remaining on treatment at month 1 and 12% at month 11. Men or women with higher education, non-migrant status, and belonging to Generation X or the Boomers, had a 66% predicted probability of remaining on treatment at month 1 and 28% at month 11, indicating longer treatment durations. While AD discontinuation may sometimes reflect remission, it is also shaped by intersectional social contexts. Our findings show that AD discontinuation is socially patterned, and this could inform comprehensive policies and programs aimed at supporting individuals from social strata that may experience more difficulties in sustaining treatment.