Abstract

Tests of hepatic function do not correlate highly and this, together with their skew distribution, makes the information provided by them difficult to interpret. By means of an appropriate transformation it is possible to normalise their distributions so that the information from them can be combined in the form of a discriminant function which is not only easy to interpret but gives an index of severity of disease which is more reliable than any one single test. In this study nine tests were examined, and after normalisation it was found that only four tests were required to present almost all the information available in the nine. A formula is given for calculating the index in T-score form, which is particularly easy to interpret. A discriminant function of the type described retains the maximum information with the minimum number of biochemical tests. It therefore contributes to the solution of a problem which is causing increasing concern--the exponential increase in the number of investigations which the hospital services will soon find it impossible to bear.