How do self and proxy dependency evaluations agree? Results from a large cohort of older women.

Background: epidemiological studies based on questionnaires can face difficulties when collecting data on functional ability of older people, and may thus use a proxy to obtain or confirm data. It is therefore of importance to compare data from older people and from their proxies, to establish to what extent proxies' answers can be used as a substitute or a complement in epidemiological studies on dependency.

Objective: we evaluated agreement in dependency assessment, and identified factors associated with discrepancy between older people and proxy reports, in a large postal epidemiological survey.

Methodology: autonomy of 5,164 community-dwelling women from the French E3N cohort study aged 75-83 was self- and proxy-evaluated with the instrumental activities of daily living (IADL) scales. Agreement was assessed using Cohen's Kappa coefficients, and factors associated with discrepancy were determined using logistic regression models.

Results: there was agreement between self and proxy dependency assessment in 90.8% of the women, with satisfactory Cohen's kappa. Discrepancy was associated with older age, poor self-reported health, prior low level of physical activity and the proxy being the partner.

Discussion: we found that a proxy could be a reliable source of dependency assessment in large epidemiological surveys, and that some characteristics of the older person and of the proxy could modulate the agreement between proxy- and self-dependency assessments.