The OBJECTIVE was to develop a methodology for predicting death in patients with burn injury using regression analysis methods. METHODS AND MATERIALS. The analysis of the results of treatment of 330 burned with a shock injury, hospitalized in the Department of Anesthesiology and Resuscitation of the Department of Thermal Lesions of Saint-Petersburg I. I. Dzhanelidze research institute of emergency medicine in the period 2013-2019. RESULTS. In the course of the study, 52 indicators were identified that characterized the condition of the victim with burn injury in the dynamics of treatment measures. To build a predictive model, only statistically significant parameters (p<0.05) were used, which were used to build a model of logistic regression. The final algorithm included 18 predictors. The model allows predicting a positive outcome of treatment and the likelihood of a fatal outcome with an accuracy of 93 and 87 % respectively. CONCLUSION. The use of a multivariate mathematical model made it possible to develop a method for predicting a fatal outcome, taking into account the peculiarities of the pathogenesis of burn disease and the principles of therapeu tic measures in the first three days after injury. The use of linear regression analysis using new indicators of thermal injury in a retrospective cohort of 330 patients allowed us to achieve a high predictive value.