Abstract
Subjects and Methods: We retrospectively analyzed 427 brain injured patients (head trauma plus cerebrovascular diseases) using a multiple logistic regression analysis to determine the factors predicting the mortality of these patients. Factors analyzed were age and sex of the patients, vital signs involving Glasgow Coma Scale (GCS), and basic laboratory studies. An optimal discriminant equation was calculated from all these parameters (EQ1) and then the equation was simplified to which only laboratory data were applied (EQ2). The areas under receiver operating characteristics (ROC) curves were compared between any two of the following four prognostic parameters: the scores by EQ1 and by EQ2, GCS score, and APACHE II score.
Result: An optimal discriminant equation (EQ1) was derived as {P=1/(1+e-x), X=8.67+0.0092(age)-0.00375 (sex;man=0, woman=1)-0.227(GCS)+0.0504(BUN)-0.700(K)-0.0451(PaCO2)+5.62×10-5(WBC)-0.0756(BE)}. The score calculated from data on admission by EQ1 was the best discriminant for all patients but was not significantly different from any one of the others. EQ2 was significantly worse predictor in the case of head trauma than any one of the others and, on the other hand, GCS was worse than others in the case of cerebrovascular diseases. Data obtained on the 7th day in ICU (n=233) was more contributory to predict prognosis than data on admission.
Conclusions: GCS was a useful predictor of mortality in patients with head trauma, while basic blood test was proved to be useful for predicting mortality of cerebrovascular diseases.
