Abstract
On the follow-up study of about 3, 000 persons who have had a set of thorough examinations, it was clarified that the cerebrovascular accident (CVA) occurred in 18 cases.
It is considered to be very important that the predicition of CVA will be made from the data of thorough examinathions. For this purpose, the establishment of the formula by means of digital computer was required to differentiate two groups with or without the high probability of the occurrence of CVA (CVA-group and non-CVA-group)
In the consideration of the following three conditions:
1) Multiple items of the examinations had to be dealt with.
2) Almost all informations from the examinations were expressed as the serial numbers.
3) The discrimination was restricted to two groups.
The application of the linear discriminant functions was adopted as the best way
As the internal samples which were used for making a formula, 15 subjects were selected from 18 subjects of CVA-group and 45 subjects of non-CVA-group with the same ages as the 15 subjects were also selected at random.
Ten items from all the examinations were selected according to the calculated weight of affectiveness to CVA by computer. Since the number of CVA group may be insufficient when all the 10 parameters were applied for the establishment of a formula by means of the linear discriminant functions, a set of 4 dimensional linear discriminant functions was derived from the combinations of the 10 parameters.
The formula was as follows:
F=0.00058(0.020X1+0.066X2-0.035X3+0.020X4)
+0.074X5+0.097X6+0.00082(2.04X7+0.25X8
+0.037X9+0.011X10)+0.023X11
X1……systolic blood pressure
X2……diastolic blood pressure
X3……maximum pressure of central retinal artery
X4……mimimum pressurre of central retinal artery
X5……eye ground findings
X6……ECG findings
X7……albuminuria
X8…maximum urine specific gravity in Fishberg's test
X9……fasting blood sugar
X10……serum cholesterol level
X11……length of observation (in year)
Mean discriminant value and variance of CVA-group were 13.84 and 2.142, respectively. Those of non-CVA-group were 10.69 and 1.752, respectively.
It was cnofirmed by the external samples, which had not been used to make the formula, whether the formula would give the best discrimination.
The rejection limits were calculated with 0.05 level of significance. The lower rejection limit of CVA-group was 9.96 and the upper rejection limit of non-CVA-group was 13.66.
It is interesting to observe that the 2 cases of non-CVA-group, which were misjudged, would have the attacks in future and it would be expected that this study of automatic prognostic diagnosis by means of electoronic computer contributes to the prophylaxis for CVA.
