We studied age-related changes of test results in persons aged 40 to 60 and the prediction of future test results in health care. The subjects were 184 individuals (143 men and 41 women) who have received health checkups for 10 successive years. The concentration of total cholesterol and glucose in men increased by aging (p <0.05), and the concentration of total cholesterol and uric acid in women also increased (p<0.05). Future individual test results could be estimated by using a linear regression equation of the four initial successive data. Although the probability of prediction was 37% to 97%, it could increase to 53% to 100% through the use of the six initial test results. This method is simple and useful to predict future test results. These predictions indicate that it could be possible to prevent lifestyle-related diseases.

View full abstract

A reference interval in the inspection of a clinical chemical is important as a reference scale of inspection value. Outliers are often in the data sampled from the reference sample group to set this interval. It is contradictory that outliers must be removed to know the distribution, though the distribution must be known to remove outliers. Therefore it will be effective in setting the reference interval if a method is found to remove outliers and to estimate the percent points without assuming the distribution of data. The exploratory data analysis (EDA) proposed by Tukey et al.(1977) has these characteristics. In this paper, a method of setting the reference interval by means of the EDA is proposed and compared with the Maximum Likelihood Transformation Method (MLTM) by simulation. It is found that the precision of MLTM is affected by the outliers; however, they hardly affect the precision of the proposed method. It is concluded that the proposed method is robust toward the types of distribution and the outliers; thus it is effective for setting a reference interval.

View full abstract