Estimation of subject-specific normal ranges based on some statistical models of an individual's physiological variations.

Three statistical models of the individual's physiological variations, proposed by Dr Eugene K. Harris, were applied to the long-term series of individual test results in our health control system. It was found from this study that, in most cases, the homeostatic model shows the best fit to the population level from comparing correlation coefficients between the observations and estimations made based on Harris's three models. It was also found that for most cases, the homeostatic model gives the most reliable estimates in individual's level from the comparison of the chi-square values between the observations and estimations made for the individual's successive test results. Further, for a group of members with hyperglycaemia, the proportion of individuals in whom the random-walk model produced the most accurate predictions was increased to the proportion in the normal group. In the age group under 39 years of age, the autoregressive model showed a relatively high degree of predictive success, while the homeostatic model showed a relatively low degree in comparison with the results found in other age groups. From these investigations, it was found that the physiologically normal state shows strict homeostatic stability.