Rolf Lefering1,2. 1. Institute for Research in Operative Medicine (IFOM), Medical Faculty, University Witten/Herdecke, Cologne, Germany. rolf.lefering@ifom-uni-wh.de. 2. Institute for Research in Operative Medicine (IFOM), Medical Faculty, University Witten/Herdecke, Ostmerheimer Str. 200, 51109, Cologne, Germany. rolf.lefering@ifom-uni-wh.de.
Abstract
INTRODUCTION: Trauma scores are often used for prognostication and the adjustment of mortality data. The appropriate consideration of identified prognostic factors is mandatory for a valid score with good outcome prediction properties. The Trauma Registry of the German Society for Trauma Surgery (TR-DGU) initially used the Trauma and Injury Severity Score (TRISS) but various reasons led to the development of a new scoring system, the Revised Injury Severity Classification (RISC). PATIENTS AND METHODS: A total of 2,008 severely injured patients with complete data documented in the TR-DGU during the period 1993-2000 were used to develop a new score. Patients were split into a development sample (n = 1,206) and a validation sample (n = 802). Multivariate logistic regression analysis was applied, and the results were compared with existing score systems. The quality of prediction was determined regarding discrimination (disparity, sensitivity, specificity, receiver operating characteristic [ROC] curve), precision (predicted versus observed mortality), and calibration (Hosmer-Lemeshow goodness-of-fit). RESULTS: Existing score systems (ISS, NISS, RTS, ASCOT, TRISS, Rixen) revealed areas under the ROC curve ranging from 0.767 to 0.877. The RISC combines 11 different components: age, NISS, head injury, severe pelvic injury, Glasgow Coma Scale, partial thromboplastin time (PTT), base excess, cardiac arrest, and indirect signs of bleeding (shock, mass transfusion, and low hemoglobin). The new RISC score reached significantly higher values of above 0.90 for the area under the ROC curve in both development and validation samples. Application to data from 2001 confirmed these results. CONCLUSION: Outcome prediction including initial laboratory values was able to significantly improve the ability to discriminate between survivors and nonsurvivors. The adjustment of mortality rates should be based on the best available prediction model.
INTRODUCTION:Trauma scores are often used for prognostication and the adjustment of mortality data. The appropriate consideration of identified prognostic factors is mandatory for a valid score with good outcome prediction properties. The Trauma Registry of the German Society for Trauma Surgery (TR-DGU) initially used the Trauma and Injury Severity Score (TRISS) but various reasons led to the development of a new scoring system, the Revised Injury Severity Classification (RISC). PATIENTS AND METHODS: A total of 2,008 severely injured patients with complete data documented in the TR-DGU during the period 1993-2000 were used to develop a new score. Patients were split into a development sample (n = 1,206) and a validation sample (n = 802). Multivariate logistic regression analysis was applied, and the results were compared with existing score systems. The quality of prediction was determined regarding discrimination (disparity, sensitivity, specificity, receiver operating characteristic [ROC] curve), precision (predicted versus observed mortality), and calibration (Hosmer-Lemeshow goodness-of-fit). RESULTS: Existing score systems (ISS, NISS, RTS, ASCOT, TRISS, Rixen) revealed areas under the ROC curve ranging from 0.767 to 0.877. The RISC combines 11 different components: age, NISS, head injury, severe pelvic injury, Glasgow Coma Scale, partial thromboplastin time (PTT), base excess, cardiac arrest, and indirect signs of bleeding (shock, mass transfusion, and low hemoglobin). The new RISC score reached significantly higher values of above 0.90 for the area under the ROC curve in both development and validation samples. Application to data from 2001 confirmed these results. CONCLUSION: Outcome prediction including initial laboratory values was able to significantly improve the ability to discriminate between survivors and nonsurvivors. The adjustment of mortality rates should be based on the best available prediction model.
Authors: Stefan Huber-Wagner; Rolf Lefering; Lars-Mikael Qvick; Markus Körner; Michael V Kay; Klaus-Jürgen Pfeifer; Maximilian Reiser; Wolf Mutschler; Karl-Georg Kanz Journal: Lancet Date: 2009-03-25 Impact factor: 79.321
Authors: Karl-Georg Kanz; April O Paul; Rolf Lefering; Mike V Kay; Uwe Kreimeier; Ulrich Linsenmaier; Wolf Mutschler; Stefan Huber-Wagner Journal: J Trauma Manag Outcomes Date: 2010-05-10
Authors: Thomas Lustenberger; Felix Walcher; Rolf Lefering; Uwe Schweigkofler; Hendrik Wyen; Ingo Marzi; Sebastian Wutzler Journal: World J Surg Date: 2016-12 Impact factor: 3.352