Background
Methods
Derivation of transition rates for different frailty states
Component | Definition |
---|---|
Weight loss | BMI < 18.5 kg/m2 OR > 10% weight loss since last wave |
Exhaustion | Positive answer to either CES-D question (‘everything I did was an effort’ or ‘I could not get going’ |
Low grip strength (adjusted for BMI) Males: Grip adjusted for BMI = Maximum grip – ((BMI-27.6)× 0.31) Females: Grip adjusted for BMI = Maximum grip – ((BMI-27.9)× 0.07) | Low grip strength was defined as values in the lowest quintile at wave 2, for males and females separately: M < 31.12 kg F < 17.60 kg |
Low gait speed (adjusted for height) Males: Gait speed adjusted for BMI = Gait speed – ((height-172)×0.00941) Females: Gait speed adjusted for BMI = Gait speed – ((height-158)× 0.010) | Low gait speed was defined as values in the lowest quintile at wave 2, for males and females separately. M < 0.691 m/s F < 0.619 m/s |
Low activity | M < 16.5 activity units F < 13.5 activity units |
Analyses - frailty transitions in ELSA
Analyses - interpolation of transition rates at 1 and 2 years
Analyses - calculation of sample sizes
Results
Transition probabilities from ELSA data
Wave 2 (n = 5344) | Wave 4 (n = 3088) | ||
---|---|---|---|
Mean age (years) (SD) | All | 71.3 (8.2) | 69.9 (7.2) |
With any missing data | 74.2 (9.7) | 73.9 (8.3) | |
Male sex (%) | All | 2389 (44.7) | 1368 (44.3) |
With any missing data | 299/694 (43.0) | 138/374 (36.9) | |
Weight loss / low BMI (%) | Calculable | 391/5107 (7.7) | 282/3088 (9.1) |
Missing | 237/5344 (4.4) | 0/3088 (0) | |
Low grip strength (%) | Calculable | 1049/5251 (20.0) | 689/2888 (23.9) |
Missing | 93/5344 (1.7) | 200/3088 (6.5) | |
Low gait speed (%) | Calculable | 977/4886 (20.0) | 694/2880 (24.1) |
Missing | 458/5344 (8.6) | 208/3088 (6.7) | |
Exhaustion (%) | Calculable | 1638/5314 (30.8) | 816/3077 (26.5) |
Missing | 30/5344 (0.6) | 11/3088 (0.4) | |
Low activity levels (%) | Calculable | 1196/5341 (22.4) | 709/3088 (23.0) |
Missing | 3/5344 (0.06) | 0/3088 (0) |
Fried Frailty score | Wave 2 (N = 4650) | Wave 4 (N = 2714) | |||
---|---|---|---|---|---|
Mean age (years) (SD) | 70.8 (7.9) | 69.3 (6.9) | |||
Male sex (%) | 2560 (55.1) | 1230 (45.3) | |||
0 (%) | Robust (%) | 2226 (47.9) | 1282 (47.2) | ||
1 (%) | Prefrail (%) | 1287 (27.7) | 1907 (41.0) | 747 (27.5) | 1108 (40.8) |
2 (%) | 620 (13.3) | 361 (13.3) | |||
3 (%) | Frail (%) | 349 (7.5) | 517 (11.1) | 223 (8.2) | 324 (11.9) |
4 (%) | 150 (3.2) | 85 (3.1) | |||
5 (%) | 18 (0.4) | 16 (0.6) |
Follow up frailty state at Wave 4 | ||||||
---|---|---|---|---|---|---|
Baseline state at Wave 2 | n | Robust | Prefrail | Frail | Dead | Dropped out |
Robust | 2226 | 937 (42) | 519 (23) | 51 (2) | 79 (4) | 640 (29) |
Prefrail | 1907 | 335 (18) | 546 (29) | 189 (10) | 178 (9) | 659 (35) |
Frail | 517 | 10 (2) | 43 (8) | 84 (16) | 133 (26) | 247 (48) |
Follow up frailty score at Wave 4 | |||||||||
---|---|---|---|---|---|---|---|---|---|
Baseline score at Wave 2 | n | 0 | 1 | 2 | 3 | 4 | 5 | Dead | Dropped out |
0 | 2226 | 937 (42) | 407 (18) | 112 (5) | 37 (2) | 13 (1) | 1 (0) | 79 (4) | 640 (29) |
1 | 1287 | 289 (23) | 259 (20) | 139 (11) | 83 (6) | 15 (1) | 2 (0) | 95 (7) | 405 (31) |
2 | 620 | 46 (7) | 67 (11) | 81 (13) | 65 (11) | 17 (3) | 7 (1) | 83 (13) | 254 (41) |
3 | 349 | 8 (2) | 13 (4) | 24 (7) | 30 (9) | 22 (6) | 2 (1) | 82 (23) | 168 (48) |
4 | 150 | 2 (1) | 1 (1) | 5 (3) | 8 (5) | 17 (11) | 3 (2) | 45 (30) | 69 (46) |
5 | 18 | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 1 (6) | 1 (6) | 6 (33) | 10 (56) |
Interpolated transition probabilities
Baseline state | Follow up state | 1 year (interpolated) | 2 year (interpolated) | 4 year (measured) |
---|---|---|---|---|
Assumes all dropouts are due to illness | Robust | 0.108 | 0.159 | 0.176 |
Prefrail | 0.688 | 0.494 | 0.286 | |
Frail | 0.087 | 0.113 | 0.099 | |
Dead/too ill | 0.117 | 0.234 | 0.439 | |
Assumes no dropouts are due to illness | Robust | 0.109 | 0.159 | 0.176 |
Prefrail | 0.688 | 0.494 | 0.286 | |
Frail | 0.087 | 0.113 | 0.099 | |
Dead | 0.019 | 0.044 | 0.093 | |
Lost to follow up | 0.097 | 0.190 | 0.346 | |
Assumes 50% of dropouts are due to illness | Robust | 0.109 | 0.159 | 0.176 |
Prefrail | 0.688 | 0.494 | 0.286 | |
Frail | 0.088 | 0.113 | 0.099 | |
Dead/too ill | 0.068 | 0.139 | 0.266 | |
Lost to follow up | 0.049 | 0.095 | 0.173 |
Illustrative sample size calculations
To detect relative risk reduction of: | Transition to frail (per group) | Transition to frail/dead/too ill (per group) | Ordinal (robust vs prefrail vs non frail/dead/too ill (per group) |
---|---|---|---|
20% | 3621 | 815 | 177 |
25% | 2389 | 558 | 117 |
30% | 1576 | 360 | 87 |
40% | 853 | 197 | 55 |
50% | 520 | 123 | 38 |
Measure | MCID | SD | Unadjusted (per group) | Adjusted (per group) |
---|---|---|---|---|
Fried frailty score | 0.3a | 1.2 | 352 | 264 |
4 m walk speed | 0.05 | 0.25 | 526 | 395 |
4 m walk speed | 0.10 | 0.25 | 132 | 99 |
SPPB | 0.5 | 2.5 | 526 | 395 |
SPPB | 1.0 | 2.5 | 132 | 99 |