OBJECTIVES: To establish a practical tool for the prediction of mortality in polytrauma patients. DESIGN: Secondary analysis of data from a public database. SETTING: A tertiary hospital in Switzerland. PARTICIPANTS: 2406 polytrauma patients were enrolled in this study, and the mean age was 44.4±19.9 years with 74.8% men. INTERVENTIONS: No. OUTCOME MEASURES: Logistic regression analysis was conducted to explore the relationship between early deaths and variables. Nomogram model was established based on predictive factors. Model effectiveness was assessed by its discrimination, calibration and decision curve analysis. Causal mediation analysis was used to explore the relationships among risk factors. RESULTS: Independent predictive variables analysed by logistic regression were: age, Glasgow Coma Scale (GCS), base excess (BE) and serum lactate. A nomogram model was established based on those risk factors, and the area under the curve of the nomogram for early death was 0.85, which was better than existing traumatic scorings. Besides, calibration plots and decision curve analysis demonstrated better performance than traumatic scorings with better internal validation. The effect of GCS on early death partly depended on BE and lactate. CONCLUSION: Our study found that early mortality was associated with age, GCS, BE and serum lactate on admission, and lactate was more important in early death. A simple prediction model of early mortality in polytrauma patients was developed with accessible parameters assessed on admission.