Xuxi Zhang

demographic variables including age, sex, country, education level and live situation in the  association between frailty as well as three domains of frailty and HRQoL. UNIANOVA was  adopted for interaction analyses. After applying Boneferroni correction for multiple testing ( P  = 0.05/40 = 0.001), no statistically significant interaction was found. All  P ‐values of the  interaction analyses are presented in Supplementary Table S2.   Analyses were performed with SPSS version 23.0 (IBM SPSS Statistics for Windows, Armonk,  NY: IBM Corp). A P ‐value < 0.05 was considered as statistically significant.  RESULTS  Participants characteristics  Table 1 summarizes the general characteristics of the study population. The mean age of  participants was 79.7 (SD = 5.6) years and 60.6% were female. Among the 2167 participants,  1195 (55.1%) were frail. Compared with participants who were not frail, frail participants were  older ( p  < 0.001), more often female ( p  < 0.001), more often had a secondary or lower  education level ( p  < 0.001), more often lived alone ( p  < 0.001), less often were at risk for  alcohol use ( p  < 0.001), less often did exercise more than once a week ( p  < 0.001) and more  often had multi‐morbidity ( p < 0.001).   Supplementary Table S1 shows the general characteristics distributed by domain of frailty.  Among the 2167 participants, 1173 (54.1%) were physically frail, 843 (38.9%) were  psychologically frail and 629 (29.0%) were socially frail.  Compared to persons included in the analysis (Supplementary Figure S1; n = 2167), persons  excluded due to missing information (n = 158) were more often smoker ( p  = 0.01) and had  lower MCS scores ( p = 0.001). No other significant differences were found between these two  groups.   Frailty and HRQoL  Table 2 presents the comparison of HRQoL scores among different frailty groups. Compared  with participants who were not frail, frail participants had significantly lower scores for both  PCS ( p  < 0.001) and MCS ( p  < 0.001) and the differences in physical HRQoL (d = 1.10) as well  as mental HRQoL (d = 0.98) were large.   Participants who were physically, psychologically or socially frail had significantly lower scores  for both PCS and MCS ( p < 0.001).   With respect to physical HRQoL, a large difference (d = 1.29) between physically and not  physically frail participants was observed, a small difference (d = 0.47) between  psychologically and not psychologically frail participants and a small difference (d = 0.39)  between socially and not socially frail participants.   32 Chapter 2