Background: previously, frailty indices were constructed using mostly subjective health measures. The reporting error in this type of measure can have implications on the robustness of frailty findings.
Objective: to examine whether frailty assessment differs when we construct frailty indices using solely self-reported or test-based health measures.
Design: secondary analysis of data from The Irish LongituDinal study on Ageing (TILDA).
Subjects and methods: 4,961 Irish residents (mean age: 61.9 ± 8.4; 54.2% women) over the age of 50 years who underwent a health assessment were included in this analysis. We constructed three frailty indices using 33 self-reported health measures (SRFI), 33 test-based health measures (TBFI) and all 66 measures combined (CFI). The 2-year follow-up outcomes examined were all-cause mortality, disability, hospitalisation and falls.
Results: all three indices had a right-skewed distribution, an upper limit to frailty, a non-linear increase with age, and had a dose-response relationship with adverse outcomes. Levels of frailty were lower when self-reported items were used (SRFI: 0.12 ± 0.09; TBFI: 0.17 ± 0.15; CFI: 0.14 ± 0.13). Men had slightly higher frailty index scores than women when test-based measures were used (men: 0.17 ± 0.09; women: 0.16 ± 0.10). CFI had the strongest prediction for risk of adverse outcomes (ROC: 0.64-0.81), and age was not a significant predictor when it was included in the regression model.
Conclusions: except for sex differences, characteristics of frailty are similar regardless of whether self-reported or test-based measures are used exclusively to construct a frailty index. Where available, self-reported and test-based measures should be combined when trying to identify levels of frailty.
Keywords: frailty; frailty index; older people; self-reported health measures; test-based health measures.
© The Author 2015. Published by Oxford University Press on behalf of the British Geriatrics Society. All rights reserved. For Permissions, please email: email@example.com.