On the choice of health inequality measure for the longitudinal analysis of income-related health inequalities