Objective: The purpose of this study was to determine whether the New Injury Severity Score (NISS) is a better predictor of mortality than the Injury Severity Score (ISS) in general and in subgroups according to age, penetrating trauma, and body region injured.
Methods: The study population consisted of 24,263 patients from three urban Level I trauma centers in the province of Quebec, Canada. Discrimination and calibration of NISS and ISS models were compared using receiver operator characteristic (ROC) curves and Hosmer-Lemeshow statistics.
Results: NISS showed better discrimination than ISS (area under the ROC curve = 0.827 vs. 0.819; p = 0.0006) and improved calibration (Hosmer-Leme-show = 62 vs. 112). The advantage of the NISS over the ISS was particularly evident among patients with head/neck injuries (area under the ROC curve = 0.819 vs. 0.784; p < 0.0001; Hosmer-Lemeshow = 59 vs. 350).
Conclusion: The NISS is a more accurate predictor of in-hospital death than the ISS and should be chosen over the ISS for case-mix control in trauma research, especially in certain subpopulations such as head/neck-injured patients.
From the Centre hospitalier affilié universitaire de Québec, Enfant-Jésus Hospital (A.L. L.M., N.L., J.S.S.), Laval University (A.L.), Quebec City, and McGill University (M.L., J.S.S.), Montreal, Quebec, Canada.
Submitted for publication December 30, 2002.
Accepted for publication April 16, 2003.
Supported by Fonds de la recherche en santé du Québec grant 015102.
Address for reprints: André Lavoie, PhD, Centre hospitalier affilié universitaire de Québec, Pavillon Enfant-Jésus, 1401, 18ème rue, Quebec City, Quebec, Canada; email: firstname.lastname@example.org.