Association between selective serotonin reuptake inhibitors and 28-day in-hospital mortality in critically ill patients with depression: A cohort study stratified by exposure timing

Selective serotonin reuptake inhibitors (SSRIs) have reported immunomodulatory effects. Whether SSRI exposure is associated with short-term survival among critically ill patients with comorbid depression, and whether associations differ by exposure timing, remains uncertain. We performed a retrospective cohort study using MIMIC-IV. SSRI exposure was modeled as a time-varying covariate in time-dependent Cox models with landmark analyses. Patients were categorized as pre-ICU maintenance users versus post-ICU initiators. Propensity-score methods were used to improve baseline comparability. As a sensitivity analysis among SSRI-naïve patients, we applied a sequential-trials stabilized IPTW marginal structural model incorporating lagged severity measures. Among 8033 patients, overall SSRI exposure showed no clear association with 28-day mortality. Associations differed by exposure pattern: post-ICU initiation was associated with a lower mortality signal (adjusted HR = 0.39, 95% CI 0.25–0.62; p < 0.0001), whereas maintenance use showed no survival advantage versus non-use. Post-ICU initiators also had lower mortality than maintenance users (adjusted HR = 0.35, 95% CI 0.21–0.57; p < 0.0001). Agent-specific findings were exploratory; fluoxetine showed the strongest association (HR = 0.28, 95% CI 0.12–0.67; p < 0.0001). In sequential-trials IPTW–MSM analyses, initiation was associated with lower standardized 28-day risk (8.25% vs 5.45%; RD = −2.80%). In ICU patients with comorbid depression, overall SSRI exposure was not clearly associated with 28-day mortality after time-aligned analyses. A lower mortality signal was observed for post-ICU initiation, whereas pre-ICU maintenance use showed no survival advantage. These findings are hypothesis-generating and warrant prospective confirmation. • Using the MIMIC-IV ICU cohort with comorbid depression, we evaluated the association between SSRI exposure and 28-day in-hospital mortality with explicit consideration of exposure timing. • We applied a time-dependent Cox model (SSRI as a time-varying exposure) and multiple landmark sensitivity analyses to mitigate time-related bias due to delayed initiation. • After bias-control analyses, pre-ICU maintenance use showed no consistent survival advantage; post-ICU initiation and agent-specific patterns suggested a potential association with lower mortality risk, but residual confounding and selection processes may remain, rendering the findings hypothesis-generating and in need of prospective validation.