Binary tree computed by conditional recursive partitioning by psychological parameters. The terminal nodes showed the direct impact of these variables on frailty. (a) Effect of SF-36-PCS (the physical indicator of health-related quality of life), together with BMI classes, number of daily drugs (in quartile, “Drugs”), absence or presence of caregiver (NO/YES), annual income, and level of education (School), on frailty classes (0 = robust, 1 = pre-frail, 2 = frail subjects). The variables affecting frailty are SF-36-PCS, having or not having a caregiver, education level, and the number of daily drugs. (b) Effect of SF-36-MCS (the mental indicator of the quality of life), together with BMI classes, number of daily drugs (“Drugs”), absence or presence of caregiver (NO/YES), annual income, and level of education (School), on frailty classes (0 = robust, 1 = pre-frail, 2 = frail subjects). The variables affecting frailty are SF-36-MCS, level of education, and the number of daily drugs. (c) Effect of GDS-15 (0–5 = no depression, 6–9 = mild depression, 10–13 = depression), together with sex, age groups, BMI classes, number of daily drugs (“Drugs”), absence or presence of caregiver (NO/YES), annual income, absence or presence of leisure activities (NO/YES), and level of education (School), on frailty classes (0 = robust, 1 = pre-frail, 2 = frail subjects). The variables affecting frailty are GDS-15, level of education, and age groups. (d) Effect of MMSE, together with sex, age groups, BMI classes, number of daily drugs (“Drugs”), absence or presence of caregiver (NO/YES), annual income, and absence or presence of leisure activities (NO/YES), on frailty classes (0 = robust, 1 = pre-frail, 2 = frail subjects). The variables affecting frailty are the number of daily drugs and leisure activities.