Dc. Becalick et Tj. Coats, Comparison of artificial intelligence techniques with UKTRISS for estimating probability of survival after trauma, J TRAUMA, 51(1), 2001, pp. 123-133
Background: The development of TRISS was principally a search for variables
that correlated with outcome. It is not known, however, if linear statisti
cal models provide optimal results. Artificial intelligence techniques can
answer this question and also determine the most important predictor variab
les.
Methods: An artificial neural network, using 16 anatomic and physiologic pr
edictor variables, was compared with the latest United Kingdom version of T
RISS model.
Results: Both methods were 89.6% correct, but TRISS was significantly bette
r by the area under the receiver operating characteristic curve (0.941 vs.
0.921, p < 0.001), The artificial neural network, however, was better calib
rated to the test data (Hosmer-Lemeshow statistic, 58.3 vs. 105.4), Head in
jury, age, and chest injury were the most important predictors by linear or
nonlinear methods, whereas respiration rate, heart rate, and systolic bloo
d pressure were underused,
Conclusion: Prediction using linear statistics is adequate but not optimal.
Only half the predictors have important predictive value, fewer still when
using linear classification. The strongest predictors swamp any nonlineari
ty observed in other variables.