Early pharmacologic venous thromboembolism prophylaxis and in-hospital mortality in ICU patients with intracerebral hemorrhage: a landmark, propensity-weighted cohort study

Background and purpose The optimal timing of pharmacologic venous thromboembolism (VTE) prophylaxis in intracerebral hemorrhage (ICH) is uncertain. Using a landmark design that classifies exposure only by information available at a fixed time point, we evaluated whether initiation of pharmacologic VTE prophylaxis within 48 h of ICU admission is associated with in-hospital mortality. Methods Retrospective cohort study in the Medical Information Mart for Intensive Care IV (MIMIC-IV v3.1). Adults with non-traumatic ICH admitted to the ICU (2008–2019) who were alive and in hospital at a 48-h landmark were included. Exposure was fixed at the landmark: early prophylaxis = pharmacologic prophylaxis started ≤48 h; no early prophylaxis = not started by 48 h (comprising later [delayed] initiators and never-treated patients). Time zero and eligibility were ICU admission + 48 h. Inverse probability of treatment weighting (IPTW) adjusted for 17 baseline covariates including the baseline Glasgow Coma Scale (GCS). The primary outcome was in-hospital mortality (a binary outcome, with discharge alive as the competing event); the secondary outcome was coded VTE (binary). Missing GCS was handled by multiple imputation. Results Of 2,754 ICH ICU patients, 2,407 reached the 48-h landmark and 2,380 formed the analytic cohort (early prophylaxis, n = 655; no early prophylaxis, n = 1,725) after excluding 27 receiving therapeutic anticoagulation within 48 h. In-hospital death occurred in 371 patients (early prophylaxis 10.8%, no early prophylaxis 17.4%). After IPTW (17/17 covariates standardized mean difference <0.1), early prophylaxis was associated with lower in-hospital mortality (adjusted odds ratio [OR] 0.72, 95% confidence interval [CI] 0.55–0.94; weighted risk difference −4.0%; cause-specific hazard ratio 0.65, 95% CI 0.50–0.86, p = 0.003). There was no statistically significant difference in coded VTE (adjusted OR 1.20, 95% CI 0.85–1.69; p = 0.31). The association was directionally consistent across most sensitivity analyses—including a model that treated prophylaxis initiation as a time-varying exposure to reduce immortal-time bias (HR 0.61, 95% CI 0.49–0.76) and a competing-risk cumulative-incidence analysis—although the 24-h landmark estimate was not statistically significant. Conclusion In this landmark cohort, initiation of pharmacologic VTE prophylaxis within 48 h was associated with lower in-hospital mortality without a statistically significant difference in coded VTE. Because key neuroimaging severity measures (hematoma volume, location, intraventricular extension, and expansion) were unavailable, residual confounding by indication cannot be excluded and a causal interpretation is not warranted. Randomized trials are needed.