Background and Aim. Septic encephalopathy (SE) is a common complication of severe sepsis. Increased concentrations of circulating soluble adhesion molecules are reported in septic patients. This study aimed to determine whether serum adhesion molecules are associated with SE. Methods. Seventy nontraumatic, nonsurgical adult patients with severe sepsis admitted through ER were evaluated. Serum adhesion molecules were assessed for their relationship with SE, and compared with other clinical predictors and biomarkers. Results. Twenty-three (32.8%) patients had SE. SE group had higher in-hospital mortality (40% versus 11%, P = 0.009) and their sVCAM-1, sICAM-1, and lactate levels on admission were also higher than non-SE group. By stepwise logistic regression model, sVCAM-1, age, and maximum 24-hours SOFA score were independently associated with septic encephalopathy. The AUC analysis of ROC curve of different biomarkers showed that sVCAM-1 is better to predict SE. The sVCAM-1 levels in the SE group were significantly higher than those of the non-SE group at three time periods (Days 1, 4, and 7). Conclusions. Septic encephalopathy implies higher mortality in nontraumatic, nonsurgical patients with severe sepsis. VCAM-1 level on presentation is a more powerful predictor of SE in these patients than lactate concentration and other adhesion molecules on admission.