The structure and predictive value of intrinsic capacity in a longitudinal study of ageing

Objectives: To assess the validity of the WHO concept of intrinsic capacity in a longitudinal study of ageing; to identify whether this overall measure disaggregated into biologically plausible and clinically useful subdomains; and to assess whether total capacity predicted subsequent care dependence.

Design: Structural equation modelling of biomarkers and self-reported measures in the English Longitudinal Study of Ageing including exploratory factor analysis, exploratory bi-factor analysis and confirmatory factor analysis. Longitudinal mediation and moderation analysis of incident care dependence.

Settings: Community, United Kingdom.

Participants: 2560 eligible participants aged over 60 years.

Main outcome measures: Activities of daily living (ADL) and instrumental activities of daily living (IADL).

Results: One general factor (intrinsic capacity) and five subfactors emerged: locomotor, cognitive; psychological; sensory; and 'vitality'. This structure is consistent with biological theory and the model had a good fit for the data (χ²=71.2 (df=39)). The summary score of intrinsic capacity and specific subfactors showed good construct validity. In a causal path model examining incident loss of ADL and IADL, intrinsic capacity had a direct relationship with the outcome-root mean square error of approximation (RMSEA)=0.02 (90% CI 0.001 to 0.05) and RMSEA=0.008 (90% CI0.001 to 0.03) respectively-and was a strong mediator for the effect of age, sex, wealth and education. Multimorbidity had an independent direct relationship with incident loss of ADLs but not IADLs, and also operated through intrinsic capacity. More of the indirect effect of personal characteristics on incident loss of ADLs and IADLs was mediated by intrinsic capacity than multimorbidity.

Conclusions: The WHO construct of intrinsic capacity appears to provide valuable predictive information on an individual's subsequent functioning, even after accounting for the number of multimorbidities. The proposed general factor and subdomain structure may contribute to a transformative paradigm for future research and clinical practice.