M�me conf�rence Poster :
IMPROVING THE TRISS METHODOLOGY BY RESTRUCTURING AGE CATEGORIES AND ADDING COMORBIDITIES Eric Bergeron MD1,2,3, Michel Rossignol MD1, Turner Osler MD(*)4, David Clas MD2,3, Andre Lavoie PhD2,3 1University of Montreal, 2Charles-LeMoyne Hospital, 3University of Sherbrooke, 4University of Vermont, 5Laval University, Quebec Background: The TRISS methodology has been developed to predict the probability of survival of a trauma population. Despite many criticisms, the methodology remains the most commonly used. We hypothesized that using different age categories and adding the presence of comorbidities would improve TRISS. Material and Methods: All blunt trauma patients older than 14 with ICD-9 code 850 to 959 from April 1993 to March 2001 were individually reviewed to collect the Revised Trauma Score (RTS), the Injury Severity Score (ISS), the age, the presence of eight well-defined comorbidities and the status at discharge: alive or dead. Two groups of equal size were created with stratification for age and year of trauma. Binary logistic regression was used to apply TRISS and to develop new models with the first group. The second group was used for cross-validation of the models. ISS was categorized as: 1-15; 16-24; 25-40; and 41-75. Age was categorized as: 15-64; 65-74; 75-84; and 85 and over. Comorbidity was dichotomized as the absence or the presence of any comorbidity. The first model included RTS, ISS and age. The second model added the presence of comorbidities. Discrimination and calibration were evaluated using the area under the receiver operating curve (AUROC) and the Hosmer-Lemeshow (HL) statistic respectively. Results: There were 5672 blunt trauma patients (2836/group). For the TRISS methodology, the HL was 12.66 and the AUROC was 0.871. For the first model, The HL was 20.35 and the AUROC was 0.902. For the second model, the HL was 14.95 and the AUROC was 0.918. The difference between each model and TRISS was statistically significant (p<0.05). When the second model was applied to the cross-validation group, the HL was 10.48 and the AUROC was 0.914 (p=NS). Conclusions: TRISS methodology is valid in predicting survival. However, a better categorization of age and the inclusion of comorbidities in the logistic model does significantly improve its predictive performance. Eric Bergeron MD Charles-LeMoyne Hospital Dept. of Traumatology 3120 Blvd. Taschereau Greenfield Park (Quebec) Canada J4V 2H1 Phone: (450) 466-5750 Fax: (450) 466-5768 Email: [EMAIL PROTECTED] http://www.aast.org/03abstracts/03absPoster_052.html __________________________________ Do you Yahoo!? Yahoo! SiteBuilder - Free, easy-to-use web site design software http://sitebuilder.yahoo.com --- URG-L Pour modifier votre adresse de courriel sur URG-L, envoyez un avis a [EMAIL PROTECTED] en indiquant votre nouvelle adresse ainsi que l'ancienne et le nom de la liste.
