Objective The predictors of poor prognosis in heat stroke (HS) remain unknown. This study investigated the predictive factors of poor prognosis in patients with HS. Methods Data were obtained and analyzed from the health records of patients diagnosed with heat illness at Ajou university hospital between January 2008 and December 2017. Univariate and multivariate analyses were performed to identify the independent predictors of poor prognosis. Results Thirty-six patients (median age, 54.5 years; 33 men) were included in the study. Poor prognosis was identified in 27.8% of the study population (10 patients). The levels of S100B protein, troponin I, creatinine, alanine aminotransferase, and serum lactate were statistically significant in the univariate analysis. Multiple regression analysis revealed that poor prognosis was significantly associated with an increased S100B protein level (odds ratio, 177.37; 95% confidence interval, 2.59 to 12,143.80; P=0.016). The S100B protein cutoff level for predicting poor prognosis was 0.610 µg/L (area under the curve, 0.906; 95% confidence interval, 0.00 to 1.00), with 86% sensitivity and 86% specificity. Conclusion An increased S100B protein level on emergency department admission is an independent prognostic factor of poor prognosis in patients with HS. Elevation of the S100B protein level represents a potential target for specific and prompt therapies in these patients.