マハラノビスの距離を用いた集団健診総合判定の試み
1992 年 34 巻 5 号 p. 448-456
詳細
1992 年 34 巻 5 号 p. 448-456
A method using Mahalanobis distance (D2), the general probability distance in multivariate analysis, was studied for evaluating the results of health examination. D2 of each subject was computed by a personal computer with a BASIC program, and those whose D2 values lay out of 95% confidence interval were considered to be abnormal. In the present study, the mean value and standard deviation come from a normal range of Japanese, and the correlation coefficients between two items were obtained from blood donors in the hospital. These three parameters were necessary for calculation of D2. To examine the availability of this method, the data of a health examination on 370 new employees from 18 to 25 yr in age were analyzed with 9 items, including systolic pressure (SBP), diastolic pressure (DBP), serum glutamic oxaloacetic transaminase activity (GOT), serum glutamic pyruvic transaminase activity (GPT), serum gamma glutamyl transpeptidase activity (GGTP), serum total cholesterol (TCHO), serum triglyceride (TG), number of erythrocytes (RBC), and concentration of hemoglobin (Hb). It was shown that the data distributions of GOT, GPT, GGTP, and TG were log normal, and those of the other items were normal. With this method, 39 persons were also judged as abnormal from 46 subjects diagnosed to be abnormal by doctors, and the remaining 7 were missed but could be classed to be normal, as six of them had only slightly high levels of GPT (33-39 IU/l) and/or GGTP (42-57 IU/l), and one had a slightly high SBP (146 mmHg). On the other hand, 41 subjects were selected as abnormal only according to this method of whom 29 demonstrated a significantly higher or lower value of one item, and 12 were due to the imbalance between some items. The individual who showed the unusual ratio of Hb to RBC and low pulse pressure (the difference between SBP and DBP) would be ignored when judged with a common method, but could be found to be abnormal with this method. This new method is therefore effective and valuable in evaluating of health examination data.
