Abstract

Background: Intimate partner violence (IPV) is a major public health concern strongly associated with adverse mental health outcomes, including depression. While evidence suggests a reciprocal relationship, few studies have formally investigated the direction and strength of these pathways using longitudinal causal modelling. This study examined the bidirectional relationship between depressive symptoms and severe IPV among women, evaluating whether exposure to severe IPV increases subsequent depressive symptoms and, conversely, whether baseline depressive symptoms elevate subsequent severe IPV risk.

Methods: We used longitudinal data from the Stepping Stones and Creating Futures (SSCF) cluster-randomised trial conducted among women (aged ≥18) in informal settlements in greater Durban, South Africa. The SSCF intervention integrates gender-transformative and livelihood-strengthening programming and was compared with a wait-list control arm. Depressive symptoms (CES-D scale) and severe IPV (recent physical and/or sexual violence) were measured at baseline, 12, and 24 months. We used cross-lagged panel models (CLPMs) to assess reciprocal relationships over time while accounting for stability effects and contemporaneous correlations. We conducted analyses separately by trial arm using weighted least squares estimation. Recognizing that traditional CLPMs do not isolate stable between-person differences from time-varying within-person changes, this method was treated as an exploratory step rather than a definitive causal determination.

Results: Depressive symptoms showed substantial temporal stability across visits in both arms, whereas severe IPV demonstrated moderate stability. Under the assumptions of no unmeasured confounding and correct temporal ordering, cross-lagged pathways indicated that prior depressive symptoms were associated with a higher likelihood of subsequent severe IPV. This pathway was particularly pronounced in the intervention arm, showing a statistically significant association between baseline depression and 12-month severe IPV. Conversely, the pathway from prior IPV to subsequent depressive symptoms was weaker and statistically non-significant at 24 months in both arms. We observed significant positive contemporaneous correlations at 12 months. Overall, the pathway from depressive symptoms to later IPV appeared stronger than the reverse relationship, though model fit statistics indicated room for improvement.

Conclusions: Depressive symptoms and severe IPV demonstrate high persistence over time, underscoring the need for sustained, integrated prevention. Prior depressive symptoms significantly linked to subsequent IPV victimisation suggest that targeted mental health screening could identify women at heightened risk. The 24-month weakening of this association in the intervention arm suggests that combined gender-transformative and livelihood interventions can disrupt this cycle. From a causal inference perspective, these exploratory results highlight the critical need for advanced approaches, such as marginal structural models or target trial designs, to fully separate the effects of dynamic, time-varying confounding from reciprocal feedback loops in resource-poor settings.