A total of 8642 older people participated in the baseline survey in 2011. Table 1 shows the participant characteristics and frailty status at baseline. Participants had a median age of 85.6 ±11.3 years, with a range of 50–114 years. At baseline, 2020 (23.4%), 2802 (32.4%), and 3820 (44.2%) participants were robust (FI score ≤0.1), pre-frail (0.1<FI score<0.25), and frail (FI score ≥0.25), respectively.
Table 1. Participant baseline frailty characteristics
|
Robust
|
Pre-frailty
|
Frailty
|
P value
|
Age, n (%)
|
|
|
|
|
|
60–75
|
955 (11.1)
|
145 (1.7)
|
918 (10.6)
|
<0.001
|
76–85
|
682 (7.9)
|
464 (5.4)
|
1193 (13.8)
|
|
86-94
|
294 (3.4)
|
852 (9.9)
|
1074 (12.4)
|
|
95-114
|
89 (1.0)
|
1341 (15.5)
|
635 (7.3)
|
|
Sex, n (%)
|
|
|
|
|
Male
|
1219 (14.1)
|
865 (10.0)
|
1821 (21.1)
|
<0.001
|
Female
|
801 (9.3)
|
1937 (22.4)
|
1999 (23.1)
|
|
Residence, n (%)
|
|
|
|
|
City
|
349 (4.0)
|
526 (6.1)
|
568 (6.6)
|
0.001
|
Town
|
642 (7.4)
|
839 (9.7)
|
1203 (13.9)
|
|
Rural
|
1029 (11.9)
|
1437 (16.6)
|
2049 (23.7)
|
|
Education level, n (%)
|
|
|
|
|
Illiterate
|
748 (8.7)
|
2074 (24.0)
|
2218 (25.7)
|
<0.001
|
Primary
|
871 (10.1)
|
552 (6.4)
|
1209 (14.0)
|
|
Middle
|
351 (4.1)
|
129 (1.5)
|
340 (3.9)
|
|
Higher
|
48 (0.6)
|
37 (0.4)
|
48 (0.6)
|
|
Marital status, n (%)
|
|
|
|
|
Single
|
27 (0.3)
|
18 (0.2)
|
42 (0.5)
|
<0.001
|
Married
|
1234 (14.3)
|
552 (6.4)
|
1548 (18.0)
|
|
Divorced or widowed
|
753 (8.7)
|
2220 (25.8)
|
2219 (25.8)
|
|
Economic status, n (%)
|
|
|
|
|
Poor
|
147 (1.7)
|
568 (6.6)
|
608 (7.1)
|
<0.001
|
Rich
|
497 (5.8)
|
387 (4.5)
|
619 (7.2)
|
|
Middle
|
1369 (16.0)
|
1794 (21.0)
|
2572 (30.0)
|
|
Total 8642
|
2020 (23.4)
|
2802 (32.4)
|
|
3820 (44.2)
|
|
In addition, 4458 participants died during the study period, as observed in 2018. The AUC of FI at baseline was 0.768 (95% CI: 0.758-0.778, P<0.001), whereas the AUC of frailty status was 0.537 (95% CI: 0.524-0.549, P<0.001), thereby showing a weaker prediction with mortality (Figure 1).
The hazard ratio (HR) of mortality according to the FI at baseline was 37.484 (95% CI: 30.217-46.498), P<0.001). Female sex (HR: 0.624, 95% CI: 0.584-0.666, P<0.001), living in the city (HR: 0.864, 95% CI: 0.792-0.943, P=0.001), being married and living with spouse (HR: 0.797, 95% CI: 0.736-0.864, P<0.001) were found to be protective factors, whereas ageing (HR: 1.057, 95% CI: 1.053-1.061, P<0.001) was a risk factor for mortality (Table 2)..
Table 2. Cox regression model analysis of the effect of the frailty index on mortality
|
B
|
SE
|
Wald
|
df
|
Sig.
|
Exp(B)
|
95% CI
|
Lower
|
Upper
|
Age
|
.056
|
.002
|
977.972
|
1
|
.000
|
1.057
|
1.053
|
1.061
|
Sex
|
-.472
|
.033
|
200.532
|
1
|
.000
|
.624
|
.584
|
.666
|
Residence
|
-.146
|
.045
|
10.754
|
1
|
.001
|
.864
|
.792
|
.943
|
Marital status
|
-.227
|
.041
|
30.716
|
1
|
.000
|
.797
|
.736
|
.864
|
FI_11
|
3.624
|
.110
|
1086.390
|
1
|
.000
|
37.484
|
30.217
|
46.498
|
Abbreviations: FI_11, frailty index in 2011; B, Regression coefficients; SE, standard error; df, degree of freedom.
We further classified frailty as non-frailty (FI <0.25) and frailty (FI ≥0.25), and analysed the HR for mortality in different states of frailty. The HR of mortality according to the FI was 2.209 (95% CI: 2.064-2.364, P<0.001) when the frailty status was dichotomized. The female sex, education level, being married, and living with spouse were found to be protective factors, whereas ageing was a risk factor of frailty. The HR for mortality was higher in pre-frail (HR: 3.588, 95% CI: 3.212-4.009, P<0.001) than in frail (HR: 1.820, 95% CI: 1.640-2.021, P<0.001) participants, when the frailty status was triaged as robust, pre-frailty, and frailty. The female sex, being married, and living with spouse were found to be protective factors, whereas ageing was a risk factor of frailty (Table 3).
Table 3. Cox regression model analysis of the effect of frailty status on mortality
|
B
|
SE
|
Wald
|
Df
|
Sig.
|
Exp(B)
|
95% CI
|
Lower
|
Upper
|
Non-frailty/Frailty
|
Age
|
.061
|
.002
|
1148.558
|
1
|
.000
|
1.062
|
1.059
|
1.066
|
Sex
|
-.464
|
.036
|
167.825
|
1
|
.000
|
.628
|
.586
|
.674
|
Education level
|
-.065
|
.026
|
6.101
|
1
|
.014
|
.937
|
.890
|
.987
|
Marital status
|
-.202
|
.041
|
24.172
|
1
|
.000
|
.817
|
.754
|
.885
|
Frailty
|
.792
|
.035
|
525.386
|
1
|
.000
|
2.209
|
2.064
|
2.364
|
Robust/Pre-frailty/frailty
|
Age
|
.058
|
.002
|
1077.900
|
1
|
.000
|
1.060
|
1.056
|
1.064
|
Sex
|
-.449
|
.033
|
183.795
|
1
|
.000
|
.638
|
.598
|
.681
|
Marital status
|
-.182
|
.041
|
19.724
|
1
|
.000
|
.834
|
.769
|
.903
|
Pre-Frailty
|
1.278
|
.057
|
510.629
|
1
|
.000
|
3.588
|
3.212
|
4.009
|
Frailty
|
.599
|
.053
|
126.720
|
1
|
.000
|
1.820
|
1.640
|
2.021
|
Abbreviations: B, Regression coefficients; SE, standard error; df, degree of freedom.
Due to the inconsistency of the different frailty status classifications, we reconsidered the FI as a continuous variable. We found that the curves of the FI at baseline and 7-year survival rate could be divided into two segments around an FI score of 0.5 (Figure 2), where the partial regression coefficients were 3.891 and 2.757, respectively. To further explore the effect of a unit increase in FI on the mortality risk, piecewise regression analysis was performed by segment within the FI score ranges of 0-0.5 and 0.5-1. When FI score was >0.5, the HR of mortality based on FI was 15.758 (95% CI: 3.656-67.924, P<0.001); however, when the FI score was ≤0.5, the HR was 48.944 (95% CI: 36.162-66.244, P<0.001). The female sex, living in the city, being married, and living with spouse were found to be protective factors, whereas ageing was a risk factor of frailty (Table 4).
Table 4. Piecewise Cox regression model analysis of the effect of frailty on mortality
|
B
|
SE
|
Wald
|
df
|
Sig.
|
Exp(B)
|
95% CI
|
Lower
|
Upper
|
FI_11 ≤0.5
|
|
|
|
|
|
|
|
|
Age
|
.057
|
.002
|
916.001
|
1
|
.000
|
1.059
|
1.055
|
1.063
|
Sex
|
-.491
|
.035
|
197.162
|
1
|
.000
|
.612
|
.571
|
.655
|
Residence
|
-.141
|
.048
|
8.806
|
1
|
.003
|
.868
|
.791
|
.953
|
Marital status
|
-.197
|
.043
|
21.259
|
1
|
.000
|
.821
|
.755
|
.893
|
FI
|
3.891
|
.154
|
634.752
|
1
|
.000
|
48.944
|
36.162
|
66.244
|
FI_11 >0.5
|
|
|
|
|
|
|
|
|
Age
|
.028
|
.006
|
21.437
|
1
|
.000
|
1.028
|
1.016
|
1.041
|
Sex
|
-.242
|
.115
|
4.454
|
1
|
.035
|
.785
|
.627
|
.983
|
Marital status
|
-.382
|
.146
|
6.789
|
1
|
.009
|
.683
|
.512
|
.910
|
FI
|
2.757
|
.745
|
13.681
|
1
|
.000
|
15.758
|
3.656
|
67.924
|
Abbreviations: FI_11, frailty index in 2011; B, Regression coefficients; SE, standard error; df, degree of freedom.