Early acetaminophen administration and clinical outcomes in critically ill patients with sepsis complicated by atrial fibrillation: a retrospective cohort study using the MIMIC-IV database with external validation

Sepsis is a leading cause of morbidity and mortality worldwide, atrial fibrillation (AF) frequently complicates its clinical course, conferring additional hemodynamic instability and adverse outcomes. We aimed to investigate the association between early acetaminophen use and mortality in patients with sepsis and atrial fibrillation and to assess the robustness of the findings using landmark and external validation analyses. We conducted a retrospective cohort study using the MIMIC‑IV database as the primary cohort and the eICU database as the external validation cohort. Adult patients with sepsis and AF were enrolled. Exposure was defined as acetaminophen use within 48 h after ICU admission. A 48-h landmark sensitivity analysis and additional landmark analyses at weeks 3 and 5 were performed to address potential survivorship and immortal time bias. The primary outcome was 28-day all-cause mortality, secondary outcomes included in-hospital, 60-day, and 90-day all-cause mortality. Propensity score matching (1:1) was applied to balance baseline characteristics between acetaminophen users and non-users. Robust Cox proportional hazards and conditional logistic regression models were used to estimate associations with outcomes. Subgroup analyses were also performed. A total of 7231 patients with sepsis and AF were identified, of whom 6125 received early acetaminophen administration during ICU admission. After propensity score matching, 2162 patients were included in the matched cohort (1081 per group), with adequate covariate balance. In the matched cohort, 28-day all-cause mortality was significantly lower among acetaminophen users than non-users (25.5% vs. 35.4%, p < 0.001). Acetaminophen use was independently associated with reduced 28-day mortality in both univariable (hazard ratio [HR] 0.65, 95% confidence interval [CI] 0.56–0.75) and multivariable analyses (HR 0.62, 95% CI 0.54–0.72). Similar associations were observed for in-hospital mortality (adjusted odds ratio 0.63, 95% CI 0.49–0.81), as well as 60-day and 90-day all-cause mortality. Subgroup analyses demonstrated consistent associations across most predefined strata. In the 48-h landmark sensitivity analysis, early acetaminophen use remained associated with lower subsequent 28-day mortality (adjusted HR, 0.57; 95% CI 0.49–0.65; p < 0.001). In the day-3 landmark analysis, cumulative exposure for 1–2 days and 3 days was associated with lower subsequent mortality, whereas the day-5 landmark analysis showed a significant association only for the 1–2-day exposure group. No consistent monotonic duration–response pattern was observed across the landmark analyses. External validation using 3169 patients from the eICU Collaborative Research Database further confirmed these findings. In multivariable logistic regression analyses, acetaminophen use remained significantly associated with lower in-hospital mortality (odds ratio [OR] 0.63, 95% confidence interval [CI] 0.53–0.74) and ICU mortality (OR 0.64, 95% CI 0.52–0.78). Early acetaminophen use was associated with lower 28-day all-cause mortality in sepsis patients complicated by AF. Findings were consistent across two independent databases, supporting the reliability and generalizability of results. However, the landmark analyses did not demonstrate a consistent duration–response relationship. These observational findings should be interpreted cautiously and do not establish causality, safety, or an optimal treatment regimen.