1979 年 19 巻 11 号 p. 1053-1061
Two hundred patients who were treated in the emergency departments of two Osaka hospitals were selected for this study. These patients were suffering from single, closed head injury, were admitted within 12 hours of trauma, and were deeply unconscious (30 or more points in 3-3-9 classification) for more than 24 hours after trauma. The relationship between initial clinical signs upon admission and actual outcome one month after admission were studied retrospectively by the chi-square test. The patients were classified into four categories: 1) able to feed himself; 2) unable to feed himself; 3) death other than brain death; and 4) brain death. Five initial features (level of consciousness, size of pupils, pupillary reaction to light, motor paralysis, and age) proved to correlate significantly (p<0.01) with outcome. Change of consciousness, rigidity, Babinski's sign, skull fracture, respiration type, and elapse of time after injury were not significantly correlated with outcome. The discriminant function was calculated by “multivariate analysis” using a computer to investigate the relationship among the above stated five features. The following formula was obtained.
Z=-1.52X1-0.57X2+0.12X3-0.40X4+0.40X5+0.36X6+0.65X7+0.29X8+0.75X9+0.02X10-0.65 (X1-9 = 0 or 1, X10=age)
Good prognosis (Outcome 1) could be discriminated from poor prognosis (Outcomes 2, 3, 4) by the Z value of zero. The discriminant ratio of this function was 67% in the good prognosis group and 74% in the poor prognosis group.
We recently applied this formula on 101 more patients and it proved to be correct in 80% of these new materials. Using the formula, we evaluated the results of treatment of 101 new patients compared with the 200 old patients by the F-test. We believe that this discriminant function can offer a prognostic index in individual patients. It is impossible to estimate the result of a therapy without such a reliable prognostic function.