BACKGROUND: The purpose of the present study was twofold: to summarize the characteristics of trauma patients with thoracic injury in the 2008 earthquake of Wenchuan, Sichuan, China, and to validate the accuracy of the Injury Severity Score (ISS), the New Injury Severity Score (NISS), and the Chest Injury Index (CII) in prediction of respiratory failure. METHODS: A group of 1,823 patients injured in the earthquake of Wenchuan were transferred to West China Hospital, Sichuan University, between May 12 and May 27, 2008. Of these 1,823 patients, 184 (10.1%) had suffered chest injury. They were enrolled in this cohort study. The risk factors of respiratory failure were analyzed with unconditional Logistic Regression. The predictive accuracies of ISS, NISS, and CII were compared with Receiver Operator Characteristic (ROC) curves. RESULTS: Of the group of patients with chest injuries, 156 (84.78%) had suffered injury to more than one organ, and 38 developed respiratory failure, for an incidence of 20.65%. The duration of mechanical ventilation ranged from 3 to 72 days (mean: 18.82 +/- 15.94 days). The occurrence of flail chest, pulmonary contusion, and crush syndrome was shown to be the risk factor for respiratory failure. There are statistical differences in area under the ROC curve between NISS and the other two trauma scores in prediction of respiratory failure. The best cut-off point of NISS is 24, with a sensitivity of 94.74% and specificity of 79.45%. CONCLUSIONS: Chest injury in earthquake is often accompanied with multiple injuries. The incidence of respiratory failure is high. The NISS can accurately predict the development of respiratory failure in chest injury patients. We recommend that NISS should be considered as a useful tool for trauma evaluation.