Early attempts to assess patient outcomes in trauma hospitals included
mortality reviews and expert panel chart audits. More recently, a sta
tistical methodology combining the Revised Trauma Score and Injury Sev
erity Score has been developed (TRISS). A modification of this methodo
logy-TRISS-like analysis-allows the inclusion of patients who have req
uired endotracheal intubation prior to the time of arrival at the trau
ma hospital. This study was undertaken to further improve this TRISS-l
ike methodology by developing statistical coefficients based on region
al data. It was hypothesized that his would allow the analysis to bett
er identify those hospitals with significantly better worse outcomes t
han their peers. The Comprehensive Data Set of the Ontario Trauma Regi
stry was accessed, which contains data on severely injured patients fr
om all 11 lead trauma hospitals in the province. Three years' data wer
e obtained, and checked for accuracy and completeness. Analysis was pe
rformed using the previously published coefficients. New coefficients
were then derived, using regression analysis on the Ontario patient da
ta. 5258 of 6389 files were complete and eligible for analysis. TRISS-
like analysis resulted in an expected mortality of 21.2% (1115.6/5258)
with a z score for the entire province of -14.102. Individual hospita
l scores were all negative (fewer deaths than expected), and 9/11 hosp
ital scores were <-1.96 (statistically significant). The new coefficie
nts were markedly different from those previously published, and their
application resulted in an overall z score of 0.000. Institutional sc
ores ranged from -3.309 to +4.686, with two hospitals <-1.96 and one >
+1.96. The old coefficients predicted many more deaths than occurred i
n all of the hospitals. The new coefficients proved quite accurate ove
rall in predicting outcomes, and identified one institution with signi
ficantly more deaths than would have been predicted for other hospital
s in the province. Subsequently, a fourth year's data files were obtai
ned, and used as a validation data set. The new coefficients again pro
ved more useful than the original ones. Copyright (C) 1997 Elsevier Sc
ience Ltd.