Optimal Timing of Prophylactic Anticoagulation in Traumatic Brain Injury: A Target Trial Emulation

Purpose To estimate the mortality‑minimizing time to initiate pharmacologic venous thromboembolism (VTE) prophylaxis after traumatic brain injury (TBI) in ICU patients and to identify patient phenotypes that may benefit from earlier vs later initiation. Methods We emulated a target trial in the National Trauma Data Bank (2019–2021) using a 24‑hour landmark and clone–censor–weight approach with stabilized inverse‑probability treatment and censoring weights. Exposure was first prophylaxis within prespecified windows (0–24, 24–48, 48–72, 72–96, 96–168 hours). Primary outcome was 14‑day in‑hospital mortality. To address heterogeneity, we derived 0–24‑hour clinical‑trajectory phenotypes via longitudinal feature engineering and k‑means clustering, then repeated policy comparisons within each phenotype. Results We identified 1,039,147 TBI patients in NTDB; 41.5% required ICU care. Among ICU patients, 45.4% received no pharmacologic prophylaxis, 36.6% received LMWH, 17.5% UFH, and 0.5% a Xa inhibitor; median time to first dose was 48.4 h. Initiation most often occurred at 0–24 h (38.7%) and 24–48 h (28.8%); later starts were associated with higher ISS, lower GCS, and increasing complications (DVT 1.7% → 5.1%, PE 1.0% → 1.7%, mortality 3.5% → 7.8%). LASSO indicated age, ISS, and neurologic severity as dominant mortality predictors; disseminated cancer markedly increased baseline hazard (HR 3.28, 1.08–10.02). Population hazard peaked in the first 24 h and decreased thereafter with no longer being significantly harmful at 72 hours (policy HRs 1.36 at 24 h; 1.15 at 48 h; 1.08 at 72 h). K-means admission trajectories yielded five phenotypes; clone–censor–weight emulation showed phenotype-specific optimal timing: Cluster 1 (moderate polytrauma) benefited at 72–96 h (HR 0.85, 0.76–0.99); Cluster 2 (severe TBI with ICP/OR) favored 96–168 h (HR 0.72, 0.62–0.94); Cluster 3 (hemorrhage-control) showed no consistent benefit; Cluster 4 (lower-severity) benefited beginning 48–72 h (HR 0.85, 0.76–0.99); and Cluster 5 (severe ICH with mass effect) showed no statistically significant protective window. Conclusions Anticoagulation timing after TBI should be phenotype-guided rather than uniform. Benefit clustered at 72–96 h for moderate polytrauma, 96–168 h for severe TBI with ICP/OR, and 48–72 h for lower-severity injuries, with no clear protective window in hemorrhage-control or severe ICH phenotypes. These findings support phenotype stratified anticoagulation protocols.