MODELLING THE EQ VAS USING THE QUANTILE REGRESSION IN ORDER TO EXPLAIN LIFESTYLE DETERMINANTS EFFECT ON QUALITY OF LIFE

Objective. The aim of this study is to propose a new approach for modelling the EQ-5D VAS, in order to explain the lifestyle determinants effect using the quantile regression analysis. Methods and Results. Data were collected within an epidemiological study conducted in Northern Italy, where 5256 subject aged >= 65 were recruited. A synthetic measure of quality of life was obtained using the EQ-5D questionnaire. With quantile regression analysis the calculation of a single value (conditional mean) is replaced by the computation of a whole set of numbers (the conditional quantiles) which provides a more complete picture of the underlying interrelations. In such a way, a much more complete view of the covariates effect on the location, scale and shape of response value distribution is provided. This method is especially suitable for EQ-5D VAS, because its distribution tends to be skewed to right, so that high values have disproportionate effects on the mean. The use of quantile regression allows to study the simultaneous effect of the considered variables on every VAS score conditional distribution quantiles. This is not possible with OLS regressions which in some cases (i.e. physical activity level, hypercholesterolemia,…) lose important information about relationship behaviour in the extreme tails of the dependent variable distribution. Conclusions. The quantile regression is very interesting because they can be performed for arbitrary quantiles of a distribution, and it provides a flexible tool for modelling the covariates effect on the full outcome variable distribution.