Background <p>Disturbances in serum osmolality are associated with poor prognosis in many diseases and are more likely to occur in patients with traumatic brain injury (TBI). However, studies correlating serum osmolality and patient prognosis are lacking. Therefore, this study investigated the correlation between serum osmolality and in-hospital all-cause mortality in patients with TBI based on a large sample of TBI patients from the Medical Information Mart for Intensive Care-IV (MIMIV-IV) database.</p> Methods <p>Patients were categorized into 4 groups based on serum osmolality levels and the association between serum osmolality and in-hospital all-cause mortality was assessed by constructing univariate and multivariate logistic regression analyses. Restricted cubic spline (RCS) curves were plotted to further assess nonlinear associations between study variables and outcomes. Kaplan–Meier analysis was used to assess the survival of patients in each group, and differences between groups were assessed by the log-rank test. Sensitivity analysis was used to assess whether this association was established in different populations.</p> Results <p>This study covered 1587 patients. The Q3 group had the lowest in-hospital mortality (7.6%). After fully adjusting for confounders, either lower or higher serum osmolality levels were associated with in-hospital all-cause mortality (Q1 vs. Q3: OR, 2.244 [1.333–3.857] <i>p</i> = 0.003; Q4 vs. Q3: OR, 2.160 [1.295–3.681] <i>p</i> = 0.004). The RCS curves showed a U-shaped correlation, with the inflection point located at a serum osmolality of 295.4 mmol/L level.</p> Conclusions <p>There was a U-shaped relationship between serum osmolality and in-hospital all-cause mortality in TBI patients. Patients had the lowest in-hospital mortality when serum osmolarity was maintained at 295.4 mmol/L.</p>